Articoli di riviste sul tema "Multi-output gaussian processes"

In another talk submitted to this conference, we presented an efficient new framework based on multi-outputs gaussian processes (MOGP) for the interpolation of few-groups homogenized cross-sections (HXS) inside deterministic core simulators. We indicated that this methodology authorized a principled selection of interpolation points through adaptive sampling. We here develop this idea by trying simple sampling schemes on our problem. In particular, we compare sample scoring functions with and without integration of leave-one-out errors, and obtained with single-output and multi-output gaussian process models. We test these methods on a realistic PWR assembly with gadolinium-added fuel rods, comparing them with non-adaptive supports. Results are promising, as the sampling algorithms allow to significantly reduce the size of interpolation supports with almost preserved accuracy. However, they exhibit phenomena of instability and stagnation, which calls for further investigation of the sampling dynamics and trying other scoring functions for the selection of samples.

Deterministic nuclear reactor simulators employing the prevalent two-step scheme often generate a substantial amount of intermediate data at the interface of their two subcodes, which can impede the overall performance of the software. The bulk of this data comprises “few-groups homogenized cross-sections” or HXS, which are stored as tabulated multivariate functions and interpolated inside the core simulator. A number of mathematical tools have been studied for this interpolation purpose over the years, but few meet all the challenging requirements of neutronics computation chains: extreme accuracy, low memory footprint, fast predictions… We here present a new framework to tackle this task, based on multi-outputs gaussian processes (MOGP). This machine learning model enables us to interpolate HXS’s with improved accuracy compared to the current multilinear standard, using only a fraction of its training data – meaning that the amount of required precomputation is reduced by a factor of several dozens. It also necessitates an even smaller fraction of its storage requirements, preserves its reconstruction speed, and unlocks new functionalities such as adaptive sampling and facilitated uncertainty quantification. We demonstrate the efficiency of this approach on a rich test case reproducing the VERA benchmark, proving in particular its scalability to datasets of millions of HXS.

Deep Gaussian Processes (DGPs) were proposed as an expressive Bayesian model capable of a mathematically grounded estimation of uncertainty. The expressivity of DPGs results from not only the compositional character but the distribution propagation within the hierarchy. Recently, it was pointed out that the hierarchical structure of DGP well suited modeling the multi-fidelity regression, in which one is provided sparse observations with high precision and plenty of low fidelity observations. We propose the conditional DGP model in which the latent GPs are directly supported by the fixed lower fidelity data. Then the moment matching method is applied to approximate the marginal prior of conditional DGP with a GP. The obtained effective kernels are implicit functions of the lower-fidelity data, manifesting the expressivity contributed by distribution propagation within the hierarchy. The hyperparameters are learned via optimizing the approximate marginal likelihood. Experiments with synthetic and high dimensional data show comparable performance against other multi-fidelity regression methods, variational inference, and multi-output GP. We conclude that, with the low fidelity data and the hierarchical DGP structure, the effective kernel encodes the inductive bias for true function allowing the compositional freedom.

The study of brain electrical activity (BEA) from different cognitive conditions has attracted a lot of interest in the last decade due to the high number of possible applications that could be generated from it. In this work, a discriminative framework for BEA via electroencephalography (EEG) is proposed based on multi-output Gaussian Processes (MOGPs) with a specialized spectral kernel. First, a signal segmentation stage is executed, and the channels from the EEG are used as the model outputs. Then, a novel covariance function within the MOGP known as the multispectral mixture kernel (MOSM) allows us to find and quantify the relationships between different channels. Several MOGPs are trained from different conditions grouped in bi-class problems, and the discrimination is performed based on the likelihood score of the test signals against all the models. Finally, the mean likelihood is computed to predict the correspondence of new inputs with each class’s existing models. Results show that this framework allows us to model the EEG signals adequately using generative models and allows analyzing the relationships between channels of the EEG for a particular condition. At the same time, the set of trained MOGPs is well suited to discriminate new input data.

Regression modelling has always been a key process in unlocking the relationships between independent and dependent variables that are held within data. In recent years, machine learning has uncovered new insights in many fields, providing predictions to previously unsolved problems. Generative Adversarial Networks (GANs) have been widely applied to image processing producing good results, however, these methods have not often been applied to non-image data. Seeing the powerful generative capabilities of the GANs, we explore their use, here, as a regression method. In particular, we explore the use of the Wasserstein GAN (WGAN) as a multi-output regression method. The resulting method we call Multi-Output Regression GANs (MOR-GANs) and its performance is compared to a Gaussian Process Regression method (GPR)—a commonly used non-parametric regression method that has been well tested on small datasets with noisy responses. The WGAN regression model performs well for all types of datasets and exhibits substantial improvements over the performance of the GPR for certain types of datasets, demonstrating the flexibility of the GAN as a model for regression.

Energy consumption forecasting plays an important role in energy management, conservation, and optimization in manufacturing companies. Aiming at the tin smelting process with multiple types of energy consumption and a strong coupling with energy consumption, the traditional prediction model cannot be applied to the multi-output problem. Moreover, the data collection frequency of different processes is inconsistent, resulting in few effective data samples and strong nonlinearity. In this paper, we propose a multi-kernel multi-output support vector regression model optimized based on a differential evolutionary algorithm for the prediction of multiple types of energy consumption in tin smelting. Redundant feature variables are eliminated using the distance correlation coefficient method, multi-kernel learning is introduced to improve the multi-output support vector regression model, and a differential evolutionary algorithm is used to optimize the model hyperparameters. The validity and superiority of the model was verified using the energy consumption data of a non-ferrous metal producer in Southwest China. The experimental results show that the proposed model outperformed multi-output Gaussian process regression (MGPR) and a multi-layer perceptron neural network (MLPNN) in terms of measurement capability. Finally, this paper uses a grey correlation analysis model to discuss the influencing factors on the integrated energy consumption of the tin smelting process and gives corresponding energy-saving suggestions.

Abstract. Sensitivity analysis of atmospheric models is necessary to identify the processes that lead to uncertainty in model predictions, to help understand model diversity through comparison of driving processes, and to prioritise research. Assessing the effect of parameter uncertainty in complex models is challenging and often limited by CPU constraints. Here we present a cost-effective application of variance-based sensitivity analysis to quantify the sensitivity of a 3-D global aerosol model to uncertain parameters. A Gaussian process emulator is used to estimate the model output across multi-dimensional parameter space, using information from a small number of model runs at points chosen using a Latin hypercube space-filling design. Gaussian process emulation is a Bayesian approach that uses information from the model runs along with some prior assumptions about the model behaviour to predict model output everywhere in the uncertainty space. We use the Gaussian process emulator to calculate the percentage of expected output variance explained by uncertainty in global aerosol model parameters and their interactions. To demonstrate the technique, we show examples of cloud condensation nuclei (CCN) sensitivity to 8 model parameters in polluted and remote marine environments as a function of altitude. In the polluted environment 95 % of the variance of CCN concentration is described by uncertainty in the 8 parameters (excluding their interaction effects) and is dominated by the uncertainty in the sulphur emissions, which explains 80 % of the variance. However, in the remote region parameter interaction effects become important, accounting for up to 40 % of the total variance. Some parameters are shown to have a negligible individual effect but a substantial interaction effect. Such sensitivities would not be detected in the commonly used single parameter perturbation experiments, which would therefore underpredict total uncertainty. Gaussian process emulation is shown to be an efficient and useful technique for quantifying parameter sensitivity in complex global atmospheric models.

A probabilistic approach to the torsional vibration analysis of a marine diesel engine propulsion shafting system is developed. The diesel engine and propeller torsional excitation are modeled probabilistically. The statistical properties of the resulting torsional vibratory shear stress in each element of the shafting system are determined by solving the corresponding input-output problem. The shafting system is considered as a multi-input linear system with the propeller and the cylinder torsional excitation as inputs and the torsional vibratory stress as the output. Under the assumption that the nonresonant torsional vibratory stresses are negligible compared with those in resonance, the input-output problem reduces to determining the statistical properties of the output of a multi-input, linear, time-invariant system driven by Gaussian amplitude modulated (AM) processes with equal carrier frequencies. The problem is solved in its general form by deriving an expression relating the double Fourier transform of the output autocorrelation function with the double Fourier transform of the input autocorrelation and cross-correlation functions. The probabilistic approach is applied to calculate the stress statistics in each shafting element of existing low-speed and medium-speed diesel engine propulsion shafting systems.

Accurate streamflow forecasting is crucial for effectively managing water resources, particularly in countries like Colombia, where hydroelectric power generation significantly contributes to the national energy grid. Although highly interpretable, traditional deterministic, physically-driven models often suffer from complexity and require extensive parameterization. Data-driven models like Linear Autoregressive (LAR) and Long Short-Term Memory (LSTM) networks offer simplicity and performance but cannot quantify uncertainty. This work introduces Sparse Variational Gaussian Processes (SVGPs) for forecasting streamflow contributions. The proposed SVGP model reduces computational complexity compared to traditional Gaussian Processes, making it highly scalable for large datasets. The methodology employs optimal hyperparameters and shared inducing points to capture short-term and long-term relationships among reservoirs. Training, validation, and analysis of the proposed approach consider the streamflow dataset from 23 geographically dispersed reservoirs recorded during twelve years in Colombia. Performance assessment reveals that the proposal outperforms baseline Linear Autoregressive (LAR) and Long Short-Term Memory (LSTM) models in three key aspects: adaptability to changing dynamics, provision of informative confidence intervals through Bayesian inference, and enhanced forecasting accuracy. Therefore, the SVGP-based forecasting methodology offers a scalable and interpretable solution for multi-output streamflow forecasting, thereby contributing to more effective water resource management and hydroelectric planning.

Abstract. Sensitivity analysis of atmospheric models is necessary to identify the processes that lead to uncertainty in model predictions, to help understand model diversity, and to prioritise research. Assessing the effect of parameter uncertainty in complex models is challenging and often limited by CPU constraints. Here we present a cost-effective application of variance-based sensitivity analysis to quantify the sensitivity of a 3-D global aerosol model to uncertain parameters. A Gaussian process emulator is used to estimate the model output across multi-dimensional parameter space using information from a small number of model runs at points chosen using a Latin hypercube space-filling design. Gaussian process emulation is a Bayesian approach that uses information from the model runs along with some prior assumptions about the model behaviour to predict model output everywhere in the uncertainty space. We use the Gaussian process emulator to calculate the percentage of expected output variance explained by uncertainty in global aerosol model parameters and their interactions. To demonstrate the technique, we show examples of cloud condensation nuclei (CCN) sensitivity to 8 model parameters in polluted and remote marine environments as a function of altitude. In the polluted environment 95 % of the variance of CCN concentration is described by uncertainty in the 8 parameters (excluding their interaction effects) and is dominated by the uncertainty in the sulphur emissions, which explains 80 % of the variance. However, in the remote region parameter interaction effects become important, accounting for up to 40 % of the total variance. Some parameters are shown to have a negligible individual effect but a substantial interaction effect. Such sensitivities would not be detected in the commonly used single parameter perturbation experiments, which would therefore underpredict total uncertainty. Gaussian process emulation is shown to be an efficient and useful technique for quantifying parameter sensitivity in complex global atmospheric model.

Wire and arc additive manufacturing (WAAM) is among the most promising additive manufacturing techniques for metals because it yields high productivity at low raw material costs. However, additional post-processing is required to remove redundant surface material from components manufactured by the WAAM process, and thus the productivity decreases. To increase productivity, multi-variable process parameters need to be optimized, including thermo-mechanical effects caused by high deposition rates. When the process is modeled, deposit shape and productivity are challenging to quantify due to uncertainty in multiple variables of the complicated WAAM process. Therefore, we modeled the WAAM process parameters, including uncertainties, using a Gaussian process regression (GPR) method, thus allowing us to develop a WAAM optimization model to improve both productivity and the quality of the deposit shape. The accuracy of the optimized output was verified through a close agreement with experimental values. The optimized deposited material had a wide effective area ratio, small height differences, and near 90° deposition angle, highlighting the usefulness of the GPR model approach to deposit nearly ideal material shapes.

Optical Earth Observation is often limited by weather conditions such as cloudiness. Radar sensors have the potential to overcome these limitations, however, due to the complex radar-surface interaction, the retrieving of crop biophysical variables using this technology remains an open challenge. Aiming to simultaneously benefit from the optical domain background and the all-weather imagery provided by radar systems, we propose a data fusion approach focused on the cross-correlation between radar and optical data streams. To do so, we analyzed several multiple-output Gaussian processes (MOGP) models and their ability to fuse efficiently Sentinel-1 (S1) Radar Vegetation Index (RVI) and Sentinel-2 (S2) vegetation water content (VWC) time series over a dry agri-environment in southern Argentina. MOGP models not only exploit the auto-correlations of S1 and S2 data streams independently but also the inter-channel cross-correlations. The S1 RVI and S2 VWC time series at the selected study sites being the inputs of the MOGP models proved to be closely correlated. Regarding the set of assessed models, the Convolutional Gaussian model (CONV) delivered noteworthy accurate data fusion results over winter wheat croplands belonging to the 2020 and 2021 campaigns (NRMSEwheat2020 = 16.1%; NRMSEwheat2021 = 10.1%). Posteriorly, we removed S2 observations from the S1 & S2 dataset corresponding to the complete phenological cycles of winter wheat from September to the end of December to simulate the presence of clouds in the scenes and applied the CONV model at the pixel level to reconstruct spatiotemporally-latent VWC maps. After applying the fusion strategy, the phenology of winter wheat was successfully recovered in the absence of optical data. Strong correlations were obtained between S2 VWC and S1 & S2 MOGP VWC reconstructed maps for the assessment dates (R2¯wheat−2020 = 0.95, R2¯wheat−2021 = 0.96). Altogether, the fusion of S1 SAR and S2 optical EO data streams with MOGP offers a powerful innovative approach for cropland trait monitoring over cloudy high-latitude regions.

For efficiently estimating the dynamic failure probability of the structure with random variables, stochastic processes and temporal and spatial multi-parameter, an estimation strategy is presented based on the random field transformation. The random field transformation focusing on the dynamic reliability with only one time parameter is further investigated, and it is extended to temporal and spatial multi-parameter issue, which simulates the output as multi-dimensional Gaussian random field. Also, the active learning Kriging method is used to construct the surrogate models for the mean function and auto-covariance function of performance function. After that, the temporal and spatial dynamic failure probability can be obtained by the simulation method. Although it doesn’t need to call the real performance function during the process of simulation method, it is time computationally expensive. To address this issue, the optimization algorithm procedure is established to estimate the dynamic failure probability. Several examples including an aero engine turbine disk and a cylindrical pressure vessel are introduced to illustrate the significance and effectiveness of the proposed methods for analyzing the temporal and spatial multi-parameter dynamic failure probability.

Motor Imagery (MI) classification using electroencephalography (EEG) has been extensively applied in healthcare scenarios for rehabilitation aims. EEG signal decoding is a difficult process due to its complexity and poor signal-to-noise ratio. Convolutional neural networks (CNN) have demonstrated their ability to extract time–space characteristics from EEG signals for better classification results. However, to discover dynamic correlations in these signals, CNN models must be improved. Hyperparameter choice strongly affects the robustness of CNNs. It is still challenging since the manual tuning performed by domain experts lacks the high performance needed for real-life applications. To overcome these limitations, we presented a fusion of three optimum CNN models using the Average Ensemble strategy, a method that is utilized for the first time for MI movement classification. Moreover, we adopted the Bayesian Optimization (BO) algorithm to reach the optimal hyperparameters’ values. The experimental results demonstrate that without data augmentation, our approach reached 92% accuracy, whereas Linear Discriminate Analysis, Support Vector Machine, Random Forest, Multi-Layer Perceptron, and Gaussian Naive Bayes achieved 68%, 70%, 58%, 64%, and 40% accuracy, respectively. Further, we surpassed state-of-the-art strategies on the BCI competition IV-2a multiclass MI database by a wide margin, proving the benefit of combining the output of CNN models with automated hyperparameter tuning.

Images under low-light conditions suffer from noise, blurring, and low contrast, thus limiting the precise detection of objects. For this purpose, a novel method is introduced based on convolutional neural network (CNN) dual attention unit (DAU) and selective kernel feature synthesis (SKFS) that merges with the Retinex theory-based model for the enhancement of dark images under low-light conditions. The model mentioned in this paper is a multi-scale residual block made up of several essential components equivalent to an onward convolutional neural network with a VGG16 architecture and various Gaussian convolution kernels. In addition, backpropagation optimizes most of the parameters in this model, whereas the values in conventional models depend on an artificial environment. The model was constructed using simultaneous multi-resolution convolution and dual attention processes. We performed our experiment in the Tesla T4 GPU of Google Colab using the Customized Raw Image Dataset, College Image Dataset (CID), Extreme low-light denoising dataset (ELD), and ExDark dataset. In this approach, an extended set of features is set up to learn from several scales to incorporate contextual data. An extensive performance evaluation on the four above-mentioned standard image datasets showed that MSR-MIRNeT produced standard image enhancement and denoising results with a precision of 97.33%; additionally, the PSNR/SSIM result is 29.73/0.963 which is better than previously established models (MSR, MIRNet, etc.). Furthermore, the output of the proposed model (MSR-MIRNet) shows that this model can be implemented in medical image processing, such as detecting fine scars on pelvic bone segmentation imaging, enhancing contrast for tuberculosis analysis, and being beneficial for robotic visualization in dark environments.

Abstract. Climate models, derived from process understanding, are essential tools in the study of climate change and its wide-ranging impacts. Hindcast and future simulations provide comprehensive spatiotemporal estimates of climatology that are frequently employed within the environmental sciences community, although the output can be afflicted with bias that impedes direct interpretation. Post-processing bias correction approaches utilise observational data to address this challenge, although they are typically criticised for not being physically justified and not considering uncertainty in the correction. This paper proposes a novel Bayesian bias correction framework that robustly propagates uncertainty and models underlying spatial covariance patterns. Shared latent Gaussian processes are assumed between the in situ observations and climate model output, with the aim of partially preserving the covariance structure from the climate model after bias correction, which is based on well-established physical laws. Results demonstrate added value in modelling shared generating processes under several simulated scenarios, with the most value added for the case of sparse in situ observations and smooth underlying bias. Additionally, the propagation of uncertainty to a simulated final bias-corrected time series is illustrated, which is of key importance to a range of stakeholders, such as climate scientists engaged in impact studies, decision-makers trying to understand the likelihood of particular scenarios and individuals involved in climate change adaption strategies where accurate risk assessment is required for optimal resource allocation. This paper focuses on one-dimensional simulated examples for clarity, although the code implementation is developed to also work on multi-dimensional input data, encouraging follow-on real-world application studies that will further validate performance and remaining limitations. The Bayesian framework supports uncertainty propagation under model adaptations required for specific applications, providing a flexible approach that increases the scope of data assimilation tasks more generally.

The Unscented Kalman Filter (UKF) is widely used for the state, disturbance, and parameter estimation of nonlinear dynamic systems, for which both process and measurement uncertainties are represented in a probabilistic form. Although the UKF can often be shown to be more reliable for nonlinear processes than the linearization-based Extended Kalman Filter (EKF) due to the enhanced approximation capabilities of its underlying probability distribution, it is not a priori obvious whether its strategy for selecting sigma points is sufficiently accurate to handle nonlinearities in the system dynamics and output equations. Such inaccuracies may arise for sufficiently strong nonlinearities in combination with large state, disturbance, and parameter covariances. Then, computationally more demanding approaches such as particle filters or the representation of (multi-modal) probability densities with the help of (Gaussian) mixture representations are possible ways to resolve this issue. To detect cases in a systematic manner that are not reliably handled by a standard EKF or UKF, this paper proposes the computation of outer bounds for state domains that are compatible with a certain percentage of confidence under the assumption of normally distributed states with the help of a set-based ellipsoidal calculus. The practical applicability of this approach is demonstrated for the estimation of state variables and parameters for the nonlinear dynamics of an unmanned surface vessel (USV).

Abstract Aims Assessment of minimum lumen areas in intravascular ultrasound (IVUS) pullbacks is time-consuming and demands adequately trained personnel. In this work, we introduce a novel and fully automated pipeline to segment the lumen boundary in IVUS datasets. Methods and results First, an automated gating is applied to select end-diastolic frames and bypass saw-tooth artefacts. Second, within a machine learning (ML) environment, we automatically segment the lumen boundary using a multi-frame (MF) convolutional neural network (MFCNN). Finally, we use the theory of Gaussian processes (GPs) to regress the final lumen boundary. The dataset consisted of 85 IVUS pullbacks (52 patients). The dataset was partitioned at the pullback-level using 73 pullbacks for training (20 586 frames), 6 pullbacks for validation (1692 frames), and 6 for testing (1692 frames). The degree of overlapping, between the ground truth and ML contours, median (interquartile range, IQR) systematically increased from 0.896 (0.874–0.933) for MF1 to 0.925 (0.911–0.948) for MF11. The median (IQR) of the distance error was also reduced from 3.83 (2.94–4.98)% for MF1 to 3.02 (2.25–3.95)% for MF11-GP. The corresponding median (IQR) in the lumen area error remained between 5.49 (2.50–10.50)% for MF1 and 5.12 (2.15–9.00)% for MF11-GP. The dispersion in the relative distance and area errors consistently decreased as we increased the number of frames, and also when the GP regressor was coupled to the MFCNN output. Conclusion These results demonstrate that the proposed ML approach is suitable to effectively segment the lumen boundary in IVUS scans, reducing the burden of costly and time-consuming manual delineation.

Fault detection in PV arrays and inverters is critical for ensuring maximum efficiency and performance. Artificial intelligence (AI) learning can be used to quickly identify issues, resulting in a sustainable environment with reduced downtime and maintenance costs. As the use of solar energy systems continues to grow, the need for reliable and efficient fault detection and diagnosis techniques becomes more critical. This paper presents a novel approach for fault detection in photovoltaic (PV) arrays and inverters, combining AI techniques. It integrates Elman neural network (ENN), boosted tree algorithms (BTA), multi-layer perceptron (MLP), and Gaussian processes regression (GPR) for enhanced accuracy and reliability in fault diagnosis. It leverages its strengths for the accuracy and reliability of fault diagnosis. Feature engineering-based sensitivity analysis was utilized for feature extraction. The fault detection and diagnosis were assessed using several statistical criteria including PBAIS, MAE, NSE, RMSE, and MAPE. Two intelligent learning scenarios are carried out. The first scenario is conducted for PV array fault detection with DC power (DCP) as output. The second scenario is conducted for inverter fault detection with AC power (ACP) as the output. The proposed technique is capable of detecting faults in PV arrays and inverters, providing a reliable solution for enhancing the performance and reliability of solar energy systems. A real-world solar energy dataset is used to evaluate the proposed technique with results compared to existing detection techniques and obtained results showing that it outperforms existing fault detection techniques, achieving higher accuracy and better performance. The GPR-M4 optimization justified its reliably among all the models with MAPE = 0.0393 and MAE = 0.002 for inverter fault detection, and MAPE = 0.091 and MAE = 0.000 for PV array fault detection.

Abstract Objective. Motor neuroprostheses require the identification of stimulation protocols that effectively produce desired movements. Manual search for these protocols can be very time-consuming and often leads to suboptimal solutions, as several stimulation parameters must be personalized for each subject for a variety of target motor functions. Here, we present an algorithm that efficiently tunes peripheral intraneural stimulation protocols to elicit functionally relevant distal limb movements. Approach. We developed the algorithm using Bayesian optimization (BO) with multi-output Gaussian Processes (GPs) and defined objective functions based on coordinated muscle recruitment. We applied the algorithm offline to data acquired in rats for walking control and in monkeys for hand grasping control and compared different GP models for these two systems. We then performed a preliminary online test in a monkey to experimentally validate the functionality of our method. Main results. Offline, optimal intraneural stimulation protocols for various target motor functions were rapidly identified in both experimental scenarios. Using the model that performed best, the algorithm converged to stimuli that evoked functionally consistent movements with an average number of actions equal to 20% of the search space size in both the rat and monkey animal models. Online, the algorithm quickly guided the observations to stimuli that elicited functional hand gestures, although more selective motor outputs could have been achieved by refining the objective function used. Significance. These results demonstrate that BO can reliably and efficiently automate the tuning of peripheral neurostimulation protocols, establishing a translational framework to configure peripheral motor neuroprostheses in clinical applications. The proposed method can also potentially be applied to optimize motor functions using other stimulation modalities.

Electric vertical takeoff and landing (eVTOL) aircraft have attracted tremendous attention nowadays due to their flexible maneuverability, precise control, cost efficiency, and low noise. The optimal takeoff trajectory design is a key component of cost-effective and passenger-friendly eVTOL systems. However, conventional design optimization is typically computationally prohibitive due to the adoption of high-fidelity simulation models in an iterative manner. Machine learning (ML) allows rapid decision making; however, new ML surrogate modeling architectures and strategies are still desired to address large-scale problems. Therefore, we showcase a novel regression generative adversarial network (regGAN) surrogate for fast interactive optimal takeoff trajectory predictions of eVTOL aircraft. The regGAN leverages generative adversarial network architectures for regression tasks with a combined loss function of a mean squared error (MSE) loss and an adversarial binary cross-entropy (BC) loss. Moreover, we introduce a surrogate-based inverse mapping concept into eVTOL optimal trajectory designs for the first time. In particular, an inverse-mapping surrogate takes design requirements (including design constraints and flight condition parameters) as input and directly predicts optimal trajectory designs, with no need to run design optimizations once trained. We demonstrated the regGAN on optimal takeoff trajectory designs for the Airbus A3 Vahana. The results revealed that regGAN outperformed reference surrogate strategies, including multi-output Gaussian processes and conditional generative adversarial network surrogates, by matching simulation-based ground truth with 99.6% relative testing accuracy using 1000 training samples. A parametric study showed that a regGAN surrogate with an MSE weight of one and a BC weight of 0.01 consistently achieved over 99.5% accuracy (denoting negligible predictive errors) using 400 training samples, while other regGAN models require at least 800 samples.

Abstract In this work, we measure the spectral index of Sagittarius A* (Sgr A*) between the H (1.6 μm) and K ′ (2.2 μm) broadband filters in the near-infrared (NIR), sampling over a factor ∼40 in brightness, the largest range probed to date by a factor ∼3. Sgr A*-NIR is highly variable, and studying the spectral index α (with F ν ∝ ν α ) is essential to determine the underlying emission mechanism. For example, variations in α with flux may arise from shifts in the synchrotron cutoff frequency, changes in the distribution of electrons, or multiple concurrent emission mechanisms. We investigate potential variations of α H − K ′ with flux by analyzing seven epochs (2005–2022) of Keck Observatory imaging observations from the Galactic Center Orbits Initiative. We remove the flux contribution of known sources confused with SgrA*-NIR, which can significantly impact color at faint flux levels. We interpolate between the interleaved H and K ′ observations using multi-output Gaussian processes. We introduce a flexible empirical model to quantify α variations and probe different scenarios. The observations are best fit by an α H − K ′ = − 0.50 ± 0.08 stat ± 0.17 sys that is constant from ∼1 mJy to ∼40 mJy (dereddened 2 μm flux). We find no evidence for a flux dependence of Sgr A*'s intrinsic spectral index. In particular, we rule out a model explaining NIR variability purely by shifts in the synchrotron cutoff frequency. We also constrain the presence of redder, quiescent emission from the black hole, concluding that the dereddened 2 μm flux contribution must be ≤0.3 mJy at 95% confidence level.

Abstract. There is a growing consensus that land surface models (LSMs) that simulate terrestrial biosphere exchanges of matter and energy must be better constrained with data to quantify and address their uncertainties. FLUXNET, an international network of sites that measure the land surface exchanges of carbon, water and energy using the eddy covariance technique, is a prime source of data for model improvement. Here we outline a multi-stage process for fusing LSMs with FLUXNET data to generate better models with quantifiable uncertainty. First, we describe FLUXNET data availability, and its random and systematic biases. We then introduce methods for assessing LSM model runs against FLUXNET observations in temporal and spatial domains. These assessments are a prelude to more formal model-data fusion (MDF). MDF links model to data, based on error weightings. In theory, MDF produces optimal analyses of the modelled system, but there are practical problems. We first discuss how to set model errors and initial conditions. In both cases incorrect assumptions will affect the outcome of the MDF. We then review the problem of equifinality, whereby multiple combinations of parameters can produce similar model output. Fusing multiple independent data provides a means to limit equifinality. We then show how parameter probability density functions (PDFs) from MDF can be used to interpret model process validity, and to propagate errors into model outputs. Posterior parameter distributions are a useful way to assess the success of MDF, combined with a determination of whether model residuals are Gaussian. If the MDF scheme provides evidence for temporal variation in parameters, then that is indicative of a critical missing dynamic process. A comparison of parameter PDFs generated with the same model from multiple FLUXNET sites can provide insights into the concept and validity of plant functional types (PFT) – we would expect similar parameter estimates among sites sharing a single PFT. We conclude by identifying five major model-data fusion challenges for the FLUXNET and LSM communities: 1) to determine appropriate use of current data and to explore the information gained in using longer time series; 2) to avoid confounding effects of missing process representation on parameter estimation; 3) to assimilate more data types, including those from earth observation; 4) to fully quantify uncertainties arising from data bias, model structure, and initial conditions problems; and 5) to carefully test current model concepts (e.g. PFTs) and guide development of new concepts.

Abstract. There is a growing consensus that land surface models (LSMs) that simulate terrestrial biosphere exchanges of matter and energy must be better constrained with data to quantify and address their uncertainties. FLUXNET, an international network of sites that measure the land surface exchanges of carbon, water and energy using the eddy covariance technique, is a prime source of data for model improvement. Here we outline a multi-stage process for "fusing" (i.e. linking) LSMs with FLUXNET data to generate better models with quantifiable uncertainty. First, we describe FLUXNET data availability, and its random and systematic biases. We then introduce methods for assessing LSM model runs against FLUXNET observations in temporal and spatial domains. These assessments are a prelude to more formal model-data fusion (MDF). MDF links model to data, based on error weightings. In theory, MDF produces optimal analyses of the modelled system, but there are practical problems. We first discuss how to set model errors and initial conditions. In both cases incorrect assumptions will affect the outcome of the MDF. We then review the problem of equifinality, whereby multiple combinations of parameters can produce similar model output. Fusing multiple independent and orthogonal data provides a means to limit equifinality. We then show how parameter probability density functions (PDFs) from MDF can be used to interpret model validity, and to propagate errors into model outputs. Posterior parameter distributions are a useful way to assess the success of MDF, combined with a determination of whether model residuals are Gaussian. If the MDF scheme provides evidence for temporal variation in parameters, then that is indicative of a critical missing dynamic process. A comparison of parameter PDFs generated with the same model from multiple FLUXNET sites can provide insights into the concept and validity of plant functional types (PFT) – we would expect similar parameter estimates among sites sharing a single PFT. We conclude by identifying five major model-data fusion challenges for the FLUXNET and LSM communities: (1) to determine appropriate use of current data and to explore the information gained in using longer time series; (2) to avoid confounding effects of missing process representation on parameter estimation; (3) to assimilate more data types, including those from earth observation; (4) to fully quantify uncertainties arising from data bias, model structure, and initial conditions problems; and (5) to carefully test current model concepts (e.g. PFTs) and guide development of new concepts.

To perform safe battery operation and avoid unnecessary degradation, battery models are needed. How complex, or close to reality the model should be, depends on the purpose of the model and the computational power available. The possibility for accurate parameterization before model implementation is another crucial aspect for the validity. Physics-based models have the advantage to solve for internal battery states, which includes important information to avoid accelerated degradation and to evaluate state of health. But the parameterization effort can be extensive and the computational cost too high to solve for in real time. Identifying the most essential processes and if possible, simplify the model is therefore motivated. Parameter sensitivity is a powerful tool to investigate how much certain parameters and the related processes effect the output signal, i.e. battery voltage. Under normal operating conditions the cells are rarely completely discharged, and as a result the diffusion processes show low sensitivity during operation [1]. In this work we therefore explore the utilization of a physics-based model solving for the electrolyte dynamics, rather than the solid phase diffusion commonly applied in the so-called single particle model. By neglecting the diffusion processes, the second dimension in the pseudo-two-dimensional (P2D) model can be ignored and the computational complexity simplified. The parameterization strategy and validation of the model is further discussed. We evaluate our model by comparing simulations with experimental data from batteries subjected to stationary energy storage applications. The cells are commercial 18650-type NMC/Graphite cells with 2.6 Ah capacity. Different service cycles have different needs in terms of current and voltage, see examples in Figure 1. Capturing these behaviours with a physics-based model might pose different challenges and will further affect the degradation of the cells [2]. A wide range of models exist in literature to capture degradation mechanisms during battery operation [3]. To keep the computational and parameterization effort as low as possible, our methodology is not to include more physics, but instead update the parameters to the processes already included in the model [4]. We explore this strategy comparing the behaviour from different types of services, as well as batteries where the application changes, to see the model response and analyse the battery state of health, see Figure 2. Electrochemical techniques such as differential voltage analysis and electrochemical impedance spectroscopy is included to support our conclusions. Our results highlight the value of application considerations when designing battery models, rather than only focusing on physical extreme points. We present an alternative physics-based model with the ability to capture crucial degradation phenomena and validate it against operational data. We believe our findings can be of value for smarter battery operating strategies and a greater understanding of application dependant lithium-ion battery degradation. Figure 1. Some of the application duty cycles part of the experimental study. a) Current profile peak shaving b) Current profile frequency regulation c) Voltage response peak shaving d) Voltage response frequency regulation. Figure 2. Degradation trends from the experimental study. Purple cell initially performing PS following application change to FR_high cycling a) Capacity evolution of the cycled cells b) tortuosity parameter evolution obtained from model parameter fitting c) Measured impedance response during cell reference performance tests d) Calculated charge transfer resistance from measured impedance data. References [1] M. Streb, M. Andersson, V. Löfqvist Klass, M. Klett, M. Johansson, G. Lindbergh, Investigating re-parametrization of electrochemical model-based battery management using real-world driving data, eTransportation 16 (2023). [2] M. Ohrelius, M. Berg, R. Wreland Lindström, G. Lindbergh, Lifetime Limitations in Multi-Service Battery Energy Storage Systems, Energies 16 (2023). [3] J.M. Reniers, G. Mulder, D.A. Howey, Review and Performance Comparison of Mechanical-Chemical Degradation Models for Lithium-Ion Batteries, Journal of The Electrochemical Society 166 (2019) A3189-A3200. [4] M. Streb, M. Ohrelius, A. Siddiqui, M. Klett, G. Lindbergh, Diagnosis and prognosis of battery degradation through re-evaluation and Gaussian process regression of electrochemical model parameters, Journal of Power Sources 588 (2023). Figure 1

Location fingerprinting using Received Signal Strength Indicators (RSSIs) has become a popular technique for indoor localization due to its use of existing Wi-Fi infrastructure and Wi-Fi-enabled devices. Artificial intelligence/machine learning techniques such as Deep Neural Networks (DNNs) have been adopted to make location fingerprinting more accurate and reliable for large-scale indoor localization applications. However, the success of DNNs for indoor localization depends on the availability of a large amount of pre-processed and labeled data for training, the collection of which could be time-consuming in large-scale indoor environments and even challenging during a pandemic situation like COVID-19. To address these issues in data collection, we investigate multi-dimensional RSSI data augmentation based on the Multi-Output Gaussian Process (MOGP), which, unlike the Single-Output Gaussian Process (SOGP), can exploit the correlation among the RSSIs from multiple access points in a single floor, neighboring floors, or a single building by collectively processing them. The feasibility of MOGP-based multi-dimensional RSSI data augmentation is demonstrated through experiments using the hierarchical indoor localization model based on a Recurrent Neural Network (RNN)—i.e., one of the state-of-the-art multi-building and multi-floor localization models—and the publicly available UJIIndoorLoc multi-building and multi-floor indoor localization database. The RNN model trained with the UJIIndoorLoc database augmented with the augmentation mode of “by a single building”, where an MOGP model is fitted based on the entire RSSI data of a building, outperforms the other two augmentation modes and results in the three-dimensional localization error of 8.42 m.

Abstract. The global ocean's oxygen content has declined significantly over the past several decades and is expected to continue decreasing under global warming, with far-reaching impacts on marine ecosystems and biogeochemical cycling. Determining the oxygen trend, its spatial pattern, and uncertainties from observations is fundamental to our understanding of the changing ocean environment. This study uses a suite of CMIP6 Earth system models to evaluate the biases and uncertainties in oxygen distribution and trends due to sampling sparseness. Model outputs are sub-sampled according to the spatial and temporal distribution of the historical shipboard measurements, and the data gaps are filled by a simple optimal interpolation method using Gaussian covariance with a constant e-folding length scale. Sub-sampled results are compared to full model output, revealing the biases in global and basin-wise oxygen content trends. The simple optimal interpolation underestimates the modeled global deoxygenation trends, capturing approximately two-thirds of the full model trends. The North Atlantic and subpolar North Pacific are relatively well sampled, and the simple optimal interpolation is capable of reconstructing more than 80 % of the oxygen trend in the non-eddying CMIP models. In contrast, pronounced biases are found in the equatorial oceans and the Southern Ocean, where the sampling density is relatively low. The application of the simple optimal interpolation method to the historical dataset estimated the global oxygen loss to be 1.5 % over the past 50 years. However, the ratio of the global oxygen trend between the sub-sampled and full model output has increased the estimated loss rate in the range of 1.7 % to 3.1 % over the past 50 years, which partially overlaps with previous studies. The approach taken in this study can provide a framework for the intercomparison of different statistical gap-filling methods to estimate oxygen content trends and their uncertainties due to sampling sparseness.

A new fluorimetric and colorimetric dual-mode probe, 4-bromo-2-(hydrazonomethyl) phenol (BHP) has been synthesized and successfully utilized for the recognition of Cd2+/F‒ ions in DMSO/H2O (9:1, v/v) system. The probe displays dual channel of detection via fluorescence enhancement and colorimetric changes upon binding with F‒ and Cd2+ ions respectively. The Job’s plot analysis, ESI-MS studies, Density Functional Theoretical (DFT) calculations, 1H NMR and 19F NMR titration results were confirmed and highly supported the 1:1 binding stoichiometry of the probe was complexed with Cd2+/F‒ ions. Furthermore, intracellular detection of F‒ ions in HeLa cells and fluorescence imaging analysis in Zebrafish embryos results of the probe BHP might be used to reveal their potential applications in a biological living system. Introduction The quantification and detection of toxic metal ions in diverse fields have fascinated more attention in recent years due to their prominent and significant roles in clinical diagnosis and ecological system.1–6 Besides metal ions, anions also play an exclusive role in a variety of chemical and biological processes.7–12 In earlier, analytical methods for the detection of cations/anions has required highly sophisticated and expensive instruments such as atomic absorption spectrometry, inductively coupled plasma mass spectrometry, ion sensitive electrodes, and gas and ion chromatography. Amid, fluorescent techniques have more expedient in terms of rapidness, excellent sensitivity and selectivity, low cost, easy and feasible detection. In addition, optical detection mode analysis is a more appropriate method because of their potential features such as easy handling, real-time analysis and different signal output modes.13–16 Besides, colorimetric assays are more feasible and potent tool as they provide a simple visible authentication for analyte detection in the absence of instruments and tedious techniques. In this perspective, the recent research area has been mainly focused to design the novel multi-functional fluorometric and colorimetric sensors for the detection of ions in the different environments. Cadmium (Cd2+) is one of the important hazardous heavy transition metal ions17 in the environment due its carcinogenic nature. The higher accumulation of Cd2+ ion and inhalation of Cd-dust prompts more awful health issues in human like cancer, cardiovascular diseases, kidneys and liver damage.18 Furthermore, the Cd2+ ion has more advantages in several industries such as pigments in plastics, electroplating and batteries, etc. On the other hand, fluoride ions play an ample role in dental health and in the treatment of osteoporosis.19–22 The excess of fluoride ingestion prompted severe disease in human health like gastric and kidney problems.23 In some remote areas, the high level contamination of fluoride ions in drinking water triggered bone disease such as fluorosis.24–31 Thus, to develop and synthesize novel multifunctional probe for the detection and quantification of both cations and anions is a highly anticipated and imperative task. Scheme 1. Synthesis of probe BHP Herein, we have fabricated and synthesized a novel chromogenic and fluorogenic assay based on bromo substituted salicylhydrazone moiety for the colorimetric and fluorometric detection of F‒ ions and colorimetric detection of Cd2+ ions in DMSO/H2O (9:1, v/v) system. The UV-visible and fluorescence spectral analysis of BHP with Cd2+/F‒ ions exposed an outstanding ratiometric absorbance and colorimetric responses towards F‒ ions and also showed a visible colorimetric response towards Cd2+ ions. The fluorescence enhancement of BHP with F‒ ion was highly evaluated by DFT calculations. As well, the cell viability experimental results of BHP can be used for the detection of F‒ ions in both HeLa cells and Zebrafish embryos via high content analysis system. Experimental Methods 2.1 Materials All the chemicals used in the present study were in the analytical reagent grade and solvents used were of HPLC grade. Reagents were used as such received without any further purification. Metal ions such as K+, Na+, Ca2+, Mg2+, Fe2+, Fe3+, Ag+, Zn2+, Mn2+, Cu2+, Co2+, Ni2+, Cd2+, Al3+, Cr3+, Pb2+ and Hg2+ were purchased from Merck and S.D. Fine chemicals. The anions of Cl-, Br-, I-, SCN-, CN-, H2PO4-, HSO4-, NO3-, AcO- and F- were purchased as their tetrabutylammonium salts from Sigma–Aldrich Pvt. Ltd. Absorption measurements were performed on JASCO V-630 spectrophotometer in 1 cm path length quartz cuvette with a volume of 2 mL at room temperature. Fluorescence measurements were made on a JASCO and F- 4500 Hitachi Spectrofluorimeter with excitation slit set at 5.0 nm band pass and emission at 5.0 nm band pass in 1 cm ×1 cm quartz cell. 1H and 13C NMR spectra were obtained on a Bruker 300 MHz NMR instrument with TMS as internal reference using DMSO-d6 as solvent. Standard Bruker software was used throughout. 19F NMR spectra were recorded at 293K on BRUKER 400 MHz FT-NMR spectrometers using DMSO-d6 as solvent. ElectroSpray Ionisation Mass Spectrometry (ESI-MS) analysis was performed in the positive/negative ion mode on a liquid chromatography-ion trap mass spectrometer (LCQ Fleet, Thermo Fisher Instruments Limited, US). Fluorescence microscopic imaging measurements were determined using Operetta High Content Imaging System (PerkinElmer, US) 2.2. Synthesis of (E)-4-bromo-2-(hydrazonomethyl) phenol, BHP An absolute alcoholic solution (50 ml) of 5-bromosalicylaldehyde (0.5gm, 2.49 mmol) was refluxed under hydrazine hydrate (in excess) for 5 hr and the pale yellow color solid product was collected after recrystallized with ethanol and ethyl acetate mixture (yield, 95 %). 1H NMR (300 MHz, DMSO-d6) δ (ppm): 8.92 (s, 1H), 11.89 (s, 1H), 7.53 (d, J = 8.7 Hz, 1H), 6.94 (d, J = 5.8 Hz, 1H); 13C NMR (75 MHz, DMSO-d6) δ (ppm): 161.36, 158.51, 135.84, 131.82, 120.86, 119.69, 106.72. 2.3 Photophysical analysis of BHP The optical mode analysis of BHP towards various cations/anions in DMSO/H2O (9:1, v/v) system was carried out by using absorbance and fluorescence spectroscopy. UV-visible and fluorescence analysis of BHP with cations were gauged by using their corresponding acetate salts of metal ions. Tetrabutylammonium salts of competing anions were used for the anionic sensing analysis. 2.4 Computational Studies The optimized geometrical and ground state energy level calculations of BHP were obtained by Density functional theoretical (DFT) calculations were executed using Gaussian 09 program 32 with the 6-311G basis set. The optimized geometries and the fluorescence enhancement of probe BHP complexed with Cd2+/F- ions were attained by DFT-B3LYP level theory using 6-311G and LANL2DZ basis sets. 2.5 Cytotoxicity studies HeLa cell lines were procured from the National Center for Cell Science (NCCS), Pune, India. Cell lines are kept in the Dulbecco's Modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 1% antimycotic and antibiotic solution was used in this study. The cells were kept in an incubator at 25 °C with humidified atmosphere comprising 5% of CO2 and 95% of air. HeLa cells were loaded over the wells of 96 well-culture plates with a density of 1 x 104 cells/well. After 48 h of incubation, previous DMEM medium was exchanged with new medium and BHP (dissolved in DMSO) was added in the range of 0-200 µM to all the wells and further incubated over 3h. Cytotoxicity of BHP was measured by using MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] assay. After incubation of HeLa cells with BHP, the medium was detached. Further, 100 μl of DMSO was added and the resulting formazan crystals were dissolved in DMSO. The cell viability was determined by measuring the absorbance of each well at 540-660 nm (formation of formazan) using a microplate reader. 2.6 In vivo fluorescence analysis in Zebrafish embryos The fluorescence imaging analysis was performed in four days old embryos. The embryos were seeded over F- ion alone for 2 h in the E3 medium. The E3 medium was prepared by dissolving 5.0 mM NaCl, 0.17mM KCl, 0.33mM CaCl2, 0.33mM MgSO4 ingredients in H2O (2L) and the pH 7.2 was adjusted by adding NaOH. The embryos were thoroughly washed with E3 medium. Successively, incubated embryos were sowed over 25 mM of BHP (in DMSO) solution for 3h. Further, embryos were washed again with E3 medium and fixed in 10% methyl cellulose solution for the good oriented images. The fluorescent images of BHP-F- were logged using high content screening microscopy. (Excitation wavelength of 482 nm and emission wavelength range of 500-700 nm). Results and discussion The probe, (E)-4-bromo-2-(hydrazonomethyl) phenol (BHP) has been synthesized by one step condensation between hydrazine and 5-bromosalicylaldehyde in ethanol (yield, 95 %) as shown in Scheme 1. The structure of the probe BHP was confirmed via 1H, 13C NMR analysis (Figure S1-S2, See ESI) 3.1. UV–vis spectral analysis of cations with BHP To investigate the cation sensing events of BHP towards different cations in DMSO/H2O (9:1, v/v) system by using UV-vis and fluorescence titration experiments. Initially, free probe BHP exhibited an absorption band at 367 nm and further addition of mono, di and trivalent cations such as Li+, K+, Ag+, Mn2+, Co2+, Ni2+, Cu2+, Zn2+, Fe2+, Hg2+, Na+, Mg2+, Ca2+, Pb2+, Fe3+ and Cr3+ exhibited tiny changes in absorption spectr due to their weak interaction towards BHP except Cd2+ ion as shown in Figure 1. Interestingly, upon titrated with Cd2+ ion, a new absorption band appeared at 470 nm due to the highly resonance induced charge transfer ability of bromo substituted salicyl moiety while the solution turns into dark yellow color from pale yellow. Increasing addition of Cd2+ ion results gradual reduction of both higher and lower energy bands at 367 nm and 470 nm respectively as depicted in Figure 2. Figure 1. UV-vis spectra of BHP (10 µM) with different cations (5 × 10-3 M) in DMSO/H2O (9: 1, v/v) system. Figure 2. UV-vis spectra of BHP (10 µM) with Cd2+ (0 – 100 µM) in DMSO/H2O (9: 1, v/v) system Besides, fluorescence response of probe BHP towards various cations such as Li+, K+, Ag+, Mn2+, Co2+, Ni2+, Cu2+, Zn2+, Fe2+, Hg2+, Na+, Mg2+, Ca2+, Pb2+, Fe3+ and Cr3+ including Cd2+ ion have been inspected in DMSO/H2O (9:1, v/v) system. Initially, the probe BHP displayed low intensed fluorescence band in free state. Addition of other commonly coexistent metal ions including Cd2+ ions exhibited trivial changes in fluorescence spectra. From these results, it is concluded that the probe BHP could serve as an excellent colorimetric assay for the detection of Cd2+ ions. 3.2. The sensing analysis of BHP towards anions Moreover, the anion binding attraction of BHP towards anions have been investigated in DMSO/H2O (9:1, v/v) system via both UV-visible and fluorescence spectral techniques. Initially the probe BHP showed the absorption band at 367 nm. Upon titrated with other anions such as Cl‒, Br‒, I‒, NO3‒, AcO‒, HSO4‒, H2PO4‒ and CN‒ were failed to alter the absorbance of the probe BHP except F‒ ions as shown in Figure 3a. Moreover, the incremental addition of F‒ ions (0-50 µM), the higher energy band at 367 nm was decreased along with the increment in new absorption band at 482 nm results an excellent ratiometric response. The new low energy band observed at 482 nm due to the deprotonation of–OH group present in salicyl moiety initiated by hydrogen bonding [Figure 3b]. At that affair, the solution turns into orange color from pale yellow and it was simply discerned by naked eye [Figure 4]. Besides, under identical condition, the fluorescence titration experiment of BHP was carried out in the presence of different anions. Interestingly, the probe BHP displayed low intensed fluorescence band at 601 nm and the other competing anions were failed to affect the fluorescence intensity except F‒ ions as shown in [Figure 5a]. Further, the incremental addition of F‒ ions triggers the enhancement in intensity results an excellent “turn on” fluorescence response due to the deprotonation and the inhibition of charge transfer state stimulated by resonance around the moiety [Figure 5b]. 3.3. Competitive experiments To gauge the selectivity and recognizing ability of BHP, competitive analysis was performed in the presence of varying concentration of F‒ ion (0-50 µM). Initially, the probe was treated with 5 × 10-3 M of different anions such as, CN-, I-, Br-, Cl-, NO2-, CH3COO-, H2PO4- and HSO4-. The other common competing anions were failed to bind with the probe BHP except F- ion [Figure 6 (a) and (b)]. From these observations, it is ensured that BHP could act as an excellent selective and sensitve chromogenic receptor for F- ions in real time monitoring and different biological applications. Figure 3 (a): UV-vis spectra of BHP with 5 × 10-3 M of other anions in DMSO/H2O (9: 1 v/v) system. (b) UV-visible spectra of BHP (5 µM) with F‒ (0-50 µM) in DMSO/H2O (9: 1 v/v) system. Figure 4. Naked eye detection of F‒ ions with BHP under visible light (top) and UV-lamp (bottom) and BHP with Cd2+ visible light only (bottom). Figure 5 (a): Fluorescence spectra of BHP (5µM) with 5 × 10-3 M of other anions in DMSO/H2O (9: 1, v/v) system. Excitation at 482 nm. Slit width is 5 nm. (b) Fluorescence spectra of BHP (5µM) with F‒ (0-50 µM) in DMSO/H2O (9: 1, v/v) system. Excitation at 482 nm. Slit width is 5 nm. Figure 6 (a): Selectivity analysis of F‒ ion with BHP in the presence of competing anions. Excitation at 480 nm, Slit width = 5 nm. (b) The blue bars represent the change of the fluorescence intensity of BHP with the consequent addition of other anions. The pink bars represent the addition of the competing anions to BHP. Excitation at 480 nm, Slit width = 5 nm. 3.4. Job’s plot analysis and calculation of binding constant of BHP for Cd2+/F‒ ions Furthermore, the Job’s plot [Figure 7(a) and (b)] analysis based on UV-visible and fluorescence titration experiments results confirmed the 1:1 binding stoichiometry of BHP with both Cd2+/F‒ ions respectively. To further support the binding stoichiometry of BHP with Cd2+/F‒ions, ESI-MS spectral analysis were performed. The ESI-MS spectral analysis of BHP-Cd2+/BHP-F‒ disclosed peaks at 327.45/258.28 corresponds to [BHP+Cd2++Na+]/[BHP+F‒+H++Na+] respectively (Figure S3-S4, See ESI). Furthermore, the 1:1 binding stoichiometry of BHP with F− ions was confirmed via 1H NMR titration profile (Figure 8) and 19F NMR. The deprotonation of ‒OH group present in the salicyl moiety was initiated by hydrogen bonding and the plausible binding mode of BHP with Cd2+ and F‒ ion is shown in Scheme 2. Further, the absorbance and fluorescence intensity changes of Cd2+ ions (A472 nm) and F‒ ions (A482 nm, I603 nm) were plotted against [Cd2+] and [F‒] respectively provided a good linear relationship between both BHP and Cd2+/F‒ ions (Figure S5, S6 and S7, See ESI). From absorbance and fluorescence titration profile, the binding constant values of BHP for Cd2+/F‒ ions were calculated using modified Benesi-Hildebrand method ions (Figure S8, S9 and S10, See ESI). The binding constant values of BHP with Cd2+ ions were found to be 4.26 ×10-4 M from UV-visible titration profile. Similarly, the binding constant values of BHP with F‒ ions were estimated to be 6.03 ×10-3 M / and 3.01 × 10-4 M from UV-visible and fluorescence titration profile respectively. The detection limits (LOD) of F‒ were calculated to be 0.05 nM respectively. Moreover, the LOD values of BHP signifies that the probe might be utilized for the quantitative determination of F‒ ions in environment and real system. Figure 7 (a) Job’s plot for BHP with F‒ ion. (b) Job’s plot for BHP with Cd2+ ion Scheme 2. Binding mode of BHP with Cd2+/F‒ ions 3.5. 1H NMR titrations of BHP with F- ions In addition, to confirm and highly supported the 1:1 binding stoichiometry of probe with F- ions, 1H NMR titrations was performed. Upon addition of F- ion (0.5 equiv), the proton signal corresponds to phenolic –OH group at 11.14 ppm was gradually decreased. Further, addition of 1 equiv. of F- ions to BHP showed the complete disappearance of –OH proton signal as depicted in Figure 8. Moreover, the binding stoichiometric ratio of F- ion with BHP was further supported by 19F NMR experiment. The (H2F)- signal appeared at -124.33 ppm (Figure S11-S12, See ESI) confirms the deprotonation process arose from phenolic –OH proton. Figure 8 1H NMR titration of BHP with F- (0-1equiv) in DMSO-d6 3.6. DFT calculations of BHP with Cd2+/F- ion To recognize the fluorescence enhancement of probe BHP after complexation with F-, DFT calculations were accomplished. The optimized structures of BHP, BHP-Cd2+ and BHP-F- were obtained using DFT/B3LYP-6-311G and B3LYP/LanL2DZ basis sets respectively. The frontier molecular orbital diagram obtained from optimized structure of BHP is presented in Figure 9. Upon binding with Cd2+ ion, the HOMO and LUMO are delocalized over the entire salicyl unit and their energy gap was reduced. It is noteworthy that inhibition of charge transfer in probe BHP renders the reduction of absorbance at 367 nm and 470 nm. Moreover, Complexation of F- ion to the probe BHP leads to lowering of HOMO-LUMO energy gap. In the presence of F-, HOMO and LUMO are distributed over the whole molecule of BHP. From these results, the F- ion was efficiently binded and complexed with BHP than Cd2+ ion. Figure 9. Frontier molecular orbital diagram of BHP, BHP-Cd2+and BHP-F‒ 3.7. Live cell Imaging analysis of BHP in HeLa cells / Zebrafish embryos The cell viability or cytotoxicity analysis of BHP (0–200 µM) against Human HeLa cells were performed using MTT assay. In 100 µM of BHP, cell viability was obtained as too high as 98%. (Figure S13, See ESI). Hence, the probe was sucessfully used for live cell imaging analysis of F- ions in Figure 10. Live cell fluorescence imaging analysis of BHP in HeLa cells. (a) Bright field images of HeLa cells incubated with BHP (25 µM) for 3h (b) Fluorescence merged images of HeLa cells incubated with BHP (25 µM) (c) Fluorescence image of HeLa cells incubated with BHP (25 µM) alone (d) Fluorescence image of HeLa cells incubated with BHP (25 µM) and 25 µM of F‒ ions for 1 h HeLa cells. Further, the HeLa cells were pre-treated with 25 µM of BHP alone for 3 h. Then HeLa cells were seaded with 25 µM of F- ions for 1h. In the absence of F- ions, the probe BHP exposed a weak yellow fluorescence. However, addition of F- ions to the probe BHP induced a bright orange fluorescence (Figure 10). These results endorsed that the probe BHP can be successfully utilized for the intracellular fluorescence imaging analysis of F- ions in HeLa cells. Besides, the exceptional cell viability output of BHP has been further explored in four days Zebrafish embryos. Zebrafish has positioned as a well-known vertebrate model in numerous biological applications. From this perspective, we have utilized also zebrafish embryos as a living animal model to expose the excellent imaging potential of BHP for the detection of F‒ ion in the biological environment (Figure 11) . Figure 11. Fluorescence imaging analysis of F‒ ion in 4 days old Zebrafish embryos developed with BHP and various concentrations of F‒ ion (a) bright field images of BHP (25 µM) alone, (b) fluorescence merged images of BHP and F- ion (25 µM) (c) fluorescence image of BHP (25 µM) alone (d) 25 µM of F‒ ion for 2 h continuously incubated with BHP (25 µM) for 3 h. 3.8. Evaluation of BHP with previous reports The probe BHP has valid and multi features such as single step synthesis, dual-mode recognition, turn-on fluorescence response and colorimetric change. The probe BHP displayed unique sensing property among other dual sensors. Table S1 compares the sensing performance of BHP with recently reported F‒ receptors. Amid, BHP exhibits too low limit of detection when compared with other previously reported chemoreceptors cited in table S1. Also, the limit of detection of BHP is within the range of recommended limits set by both EPA and WHO for F‒ Ions. Moreover, the fluorescence imaging experiments inferred that the probe BHP can be utilized as potential tool for mapping F‒ ion distribution in HeLa cells and Zebrafish embryos. Conclusions We have designed and synthesized a new chromogenic and fluorogenic probe based on salicylhydrazone derivative for the selective and sensitive detection of both Cd2+/F- ions by colorimetrically and fluorimetrically respectively. As per our knowledge, it is a novel simple hydrazone receptor for sensing carcinogenic heavy metal Cd2+ via colorimetric method and biologically significant F‒ ion by both colorimetric and fluorimetric methods. The binding constant value of Cd2+ ion was found to be 4.26×10-4 M by UV-visible method where as 6.03×10-3 and 3.01×10-4 M for F- ion by both UV-visible and fluorescence methods respectively. The limit of detection was found to be 0.05 nM for F- ion. The excellent biological potential of BHP has been successfully utilized for the detection of F- ions in Zebrafish embryos and human HeLa cells. Acknowledgments The authors acknowledge the financial support from the Council of Scientific and Industrial Research, Extramural Research, New Delhi, India (Grant No. 01(2901)17/EMR-II. The Department of Science and Technology, SERB, Extramural Major Research Project (Grant No. EMR/2015/000969), Department of Science and Technology, CERI, New Delhi, India (Grant No. DST/TM/CERI/C130(G) and we acknowledge the DST-FIST, DST-PURSE,DST-IRPHA, UPE programme and UGC-NRCBS, SBS, MKU for providing instrumentation facilities. References Jäkle, F. Chem. Rev. 2010, 110, 3985. Chen, X.; Zhou, Y.; Peng, X.; Yoon, J. Chem. Soc. 2010, 39, 2120. Kim, H. N.; Guo, Z.; Zhu, W.; Yoon J.; Tian, H. Chem. Soc. Rev. 2011, 40, 79. Zhang, J. F.; Zhou, Y.; Yoon, J.; Kim, J. S. Chem. Soc. Rev. 2011, 40, 3416. Zhou, Y.; Yoon, J. Chem. Soc. Rev. 2012, 41, 52. Chen, X.; Pradhan, T.; Wang, F.; Kim, J. S.; Yoon, J. Chem. Rev. 2012, 112, 1910. Wade, C. R.; Broomsgrove, A. E. J.; Aldridge, S.; Gabbaï, F. P. Chem. Rev. 2010, 110, Gale, P. A. Chem. Soc. Rev. 2010, 39, 3746. Wenzel, M.; Hiscock, J. R.; Gale P. A. Chem. Soc. Rev. 2012, 41, 480. Xu, Z.; Chen, X.; Kim, H. N.; Yoon, J. Chem. Soc. Rev. 2010, 39, 127. Galbraith, E.; James, T. D.; Chem. Soc. Rev. 2010, 39, 3831. Xu, Z.; Kim, S. K.; Yoon, J. Chem. Soc. Rev. 2010, 39, 1457. Zhou,Y.; Xu, Z.; Yoon, J.; Chem. Soc. Rev. 2011, 40, Quang, D. T.; Kim, J. S. Chem. Rev. 2010, 110, 6280. Zhou, Y.; Zhang, J. F.; Yoon, J.; Chem. Rev. 2014, 114, 5511. Martínez-Máñez, R.; Sancenón, F. Fluorogenic and chromogenic chemosensors and reagents for anions, 2003. Waisberg, M.; Joseph, P.; Hale, B.; Beyersmann, D. Toxicology. 2003, 192, 95. McFarland, C. N.; Bendell-Young, L. I.; Guglielmo, C.; Williams, T. D. J. Environ. Monit. 2002, 4, Nordberg, G. F.; Herber R. F. M.; Alessio, L. Cadmium in the Human Environment, Oxford University Press, Oxford, UK, 1992. Akula, M.; Thigulla, Y.; Nag, A.; Bhattacharya, RSC Adv. 2015, 5, 57231. Michael Kleerekoper. Clin. North Am. 2018, 65, 441. Erdal, S.; Buchanan, S. N.; Environ. Health Perspect. 2005, 113, Michigami, Y.; Kuroda, Y.; Ueda, K.; Yamamoto, Y.; Anal. Chim. 1993, 274, 299. Sivamani, J.; Siva, A. Sens. Actuators, B, 2017, 242, 423. Sarveswari, S.; Jesin Beneto, A.; Siva, A. Sens. Actuators, B, 2017, 245, 428. Sivamani, J.; Sadhasivam, V.; Siva, A.; Sens. Actuators, 2017, 246, 108. Jesin Beneto, A.; Siva, A.; Sens. Actuators, 2017, 247, 526. Krishnaveni, K.; Iniya, M.; Jeyanthi, D.; Siva, A.; Chellappa, D. Spectrochim. Acta A. 2018, 205, 557. Krishnaveni, K.; Murugesan, S.; Siva, A. New J. 2019, 4, 9021. Krishnaveni, K.; Iniya, M.; Siva, A.; Vidhyalakshmi, N.; Ramesh, U.; Sasikumar, S.; Murugesan, S.; J. Mol. 2020, 1217, 128446. Krishnaveni, K.; Harikrishnan, M.; Murugesan S.; Siva, A. Microchem. J. 2020, 159, 105543. Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Scalmani, G.; Barone, B. Mennucci, G. A. Petersson, H. Nakatsuji, M. Caricato, V.; Li, X.; Hratchian, H. P.; Izmaylov, A. F.; Bloino, J.; Zheng, G.; Sonnenberg, J. L.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Vreven, T.; Montgomery, J. A.; Peralta, J. E. Jr.; Ogliaro, F.; Bearpark, M.; Heyd, J. J.; Brothers, E.; Kudin, K. N.; Staroverov, V. N.; Kobayashi, R.; Normand, J.; Raghavachari, K.; Rendell, A.; Burant, J. C.; Iyengar, S. S.; Tomasi, J.; Cossi, M.; Rega, N.; Millam, J. M.; Klene, M.; Knox, J. E.; Cross, J. B.; Bakken, V.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Martin, R. L.; Morokuma, K.; Zakrzewski, V. G.; Voth, A.; Salvador, P.; Dannenberg, J. J.; Dapprich, S.; Daniels, A. D.; Farkas, O.; Foresman, J. B.; Ortiz, J. V.; Cioslowski J.; Fox, D. J. Gaussian 09, Revision A. 02, Gaussian, Inc., Wallingford C T, York, 1991.

PurposeA recent innovative technology used in wireless communication is recognized as multiple input multiple output (MIMO) communication system and became popular for quicker data transmission speed. This technology is being examined and implemented for the latest broadband wireless connectivity networks. Though high-capacity wireless channel is identified, there is still requirement of better techniques to get increased data transmission speed with acceptable reliability. There are two types of systems comprising of multi-antennas placed at transmitting and receiving sides, of which first is diversity technique and another is spatial multiplexing method. By making use of these diversity techniques, the reliability of transmitting signal can be improved. The fundamental method of the diversity is to transform wireless channel such as Rayleigh fading into steady additive white Gaussian noise (AWGN) channel which is devoid of any disastrous fading of the signal. The maximum transmission speed that can be achieved by spatial multiplexing methods is nearly equal to channel capacity of MIMO. Conversely, for diversity methods, the maximum speed of broadcasting is much lower than channel capacity of MIMO. With the advent of space–time block coding (STBC) antenna diversity technique, higher-speed data transmission is achievable for spatially multiplexed multiple input multiple output (SM-MIMO) system. At the receiving end, detection of the signal is a complex task for system which exhibits SM-MIMO. Additionally, a link modification method is implemented to decide appropriate coding and modulation scheme such as space diversity technique STBC to use two-way radio resources efficiently. The proposed work attempts to improve detection of signal at receiving end by employing STBC diversity technique for linear detection methods such as zero forcing (ZF), minimum mean square error (MMSE), ordered successive interference cancellation (OSIC) and maximum likelihood detection (MLD). The performance of MLD has been found to be better than other detection techniques.Design/methodology/approachAlamouti's STBC uses two transmit antennas regardless of the number of receiver antennas. The encoding and decoding operation of STBC is shown in the earlier cited diagram. In the following matrix, the rows of each coding scheme represent a different time instant, while the columns represent the transmitted symbols through each different antenna. In this case, the first and second rows represent the transmission at the first and second time instant, respectively. At a time t, the symbol s1 and symbol s2 are transmitted from antenna 1 and antenna 2, respectively. Assuming that each symbol has duration T, then at time t + T, the symbols –s2* and s1*, where (.)* denotes the complex conjugate, are transmitted from antenna 1 and antenna 2, respectively. Case of one receiver antenna: The reception and decoding of the signal depend on the number of receiver antennas available. For the case of one receiver antenna, the received signals are received at antenna 1 , hij is the channel transfer function from the jth transmit antenna and the ith receiver antenna, n1 is a complex random variable representing noise at antenna 1 and x (k) denotes x at time instant k ( at time t + (k – 1)T.FindingsThe results obtained for maximal ratio combining (MRC) with 1 × 4 scheme show that the BER curve drops to 10–4 for signal-to-noise (SNR) ratio of 10 dB, whereas for MRC 1 × 2 scheme, the BER drops down to 10–5 for SNR of 20 dB. Results obtained in Table 1 show that when STBC is employed for MRC with 1 × 2 scheme (one antenna at transmitter node and two antennas at receiver node), BER curve comes down to 0.0076 for Eb/N0 of 12. Similarly, when MRC with 1 × 4 antenna scheme is implemented, BER drops down to 0 for Eb/N0 of 12. Thus, it can be concluded from the obtained graph that the performance of MRC with STBC gives improved results. When STBC technique is used with 3 × 4 scheme, at SNR of 10 dB, BER comes nearer to 10–6 (figure 7.3). It can be concluded from the analytics observed between AWGN and Rayleigh fading channel that for AWGN channel, BER is found to be equal to 0 for SNR value of 13.5 dB, whereas for Rayleigh fading channel, BER is observed nearer to 10–3 for Eb/N0 = 15. Simulation results (in figure 7.2) from the analytics show BER drops to 0 for SNR value of 12 dB.Research limitations/implicationsOptimal design and successful deployment of high-performance wireless networks present a number of technical challenges. These include regulatory limits on useable radio-frequency spectrum and a complex time-varying propagation environment affected by fading and multipath. The effect of multipath fading in wireless systems can be reduced by using antenna diversity. Previous studies show the performance of transmit diversity with narrowband signals using linear equalization, decision feedback equalization, maximum likelihood sequence estimation (MLSE) and spread spectrum signals using a RAKE receiver. The available IC techniques compatible with STBC schemes at transmission require multiple antennas at the receiver. However, if this not a strong constraint at the base station level, it remains a challenge at the handset level due to cost and size limitation. For this reason, SAIC technique, alternative to complex ML multiuser demodulation technique, is still of interest for 4G wireless networks using the MIMO technology and STBC in particular. In a system with characteristics similar to the North American Digital mobile radio standard IS-54 (24.3 K symbols per sec. with an 81 Hz fading rate), adaptive retransmission with time deviation is not practical.Practical implicationsThe evaluation of performance in terms of bit error rate and convergence time which estimates that MLD technique outperforms in terms of received SNR and low decoding complexity. MLD technique performs well but when higher number of antennas are used, it requires more computational time and thereby resulting in increased hardware complexity. When MRC scheme is implemented for singe input single output (SISO) system, BER drops down to 10–2 for SNR of 20 dB. Therefore, when MIMO systems are employed for MRC scheme, improved results based on BER versus SNR are obtained and are used for detecting the signal; comparative study based on different techniques is done. Initially ZF detection method is utilized which was then modified to ZF with successive interference cancellation (ZFSIC). When successive interference cancellation scheme is employed for ZFSIC, better performance is observed as compared to the estimation of ML and MMSE. For 2 × 2 scheme with QPSK modulation method, ZFSIC requires more computational time as compared to ZF, MMSE and ML technique. From the obtained results, the conclusion is that ZFSIC gives the improved results as compared to ZF in terms of BER ratio. ZF-based decision statistics can be produced by the detection algorithm for a desired sub-stream from the received vector whichs consist of an interference which occurred from previous transmitted sub-streams. Consequently, a decision on the secondary stream is made and contribution of the noise is regenerated and subtracted from the vector received. With no involvement of interference cancellation, system performance gets reduced but computational cost is saved. While using cancellation, as H is deflated, coefficients of MMSE are recalculated at each iteration. When cancellation is not involved, the computation of MMSE coefficients is done only once, because of H remaining unchanged. For MMSE 4 × 4 BPSK scheme, bit error rate of 10–2 at 30 dB is observed. In general, the most thorough procedure of the detection algorithm is the computation of the MMSE coefficients. Complexity arises in the calculation of the MMSE coefficients, when the antennas at the transmitting side are increased. However, while implementing adaptive MMSE receivers on slow channel fading, it is probable to recover the signal with the complications being linear in the antennas of transmitter node. The performance of MMSE and successive interference cancellation of MMSE are observed for 2 × 2 and 4 × 4 BPSK and QPSK modulation schemes. The drawback of MMSE SIC scheme is that the first detected signal observes the noise interference from (NT-1) signals, while signals processed from every antenna later observe less noisy interference as the process of cancellation progresses. This difficulty could be overcome by using OSIC detection method which uses successive ordering of the processed layers in the decreasing power of the signal or by power allocation to the signal transmitted depending on the order of the processing. By using successive scheme, a computation of NT delay stages is desired to bring out the abandoned process. The work also includes comparison of BER with various modulation schemes and number of antennas involved while evaluating the performance. MLD determines the Euclidean distance among the vector signal received and result of all probable transmitted vector signals with the specified channel H and finds the one with the minimum distance. Estimated results show that higher order of the diversity is observed by employing more antennas at both the receiving and transmitting ends. MLD with 8 × 8 binary phase shift keying (BPSK) scheme offers bit error rate near to 10–4 for SNR (16 dB). By using Altamonti space ti.Social implicationsIt should come as no surprise that companies everywhere are pushing to get products to market faster. Missing a market window or a design cycle can be a major setback in a competitive environment. It should be equally clear that this pressure is coming at the same time that companies are pushing towards “leaner” organizations that can do more with less. The trends mentioned earlier are not well supported by current test and measurement equipment, given this increasingly high-pressure design environment: in order to measure signals across multiple domains, multiple pieces of measurement equipment are needed, increasing capital or rental expenses. The methods available for making cross-domain, time-correlated measurements are inefficient, reducing engineering efficiency. When only used on occasion, the learning curve to understand how to use equipment for logic analysis, time domain and RF spectrum measurements often requires an operator to re-learn each piece of separate equipment. The equipment needed to measure wide bandwidth, time-varying spectral signals is expensive, again increasing capital or rental expenses. What is needed is a measurement instrument with a common user interface that integrates multiple measurement capabilities into a single cost-effective tool that can efficiently measure signals in the current wide-bandwidth, time-correlated, cross-domain environments. The market of wireless communication using STBCs has large scope of expansion in India. Therefore, the proposed work has techno-commercial potential and the product can be patented. This project shall in turn be helpful for remote areas of the nearby region particularly in Gadchiroli district and Melghat Tiger reserve project of Amravati district, Nagjira and so on where electricity is not available and there is an all the time problem of coverage in getting the network. In some regions where electricity is available, the shortage is such that they cannot use it for peak hours. In such cases, stand-alone space diversity technique, STBC shall help them to meet their requirements in making connection during coverage problem, thereby giving higher data transmission rates with better QOS (quality of service) with least dropped connections. This trend towards wireless everywhere is causing a profound change in the responsibilities of embedded designers as they struggle to incorporate unfamiliar RF technology into their designs. Embedded designers frequently find themselves needing to solve problems without the proper equipment needed to perform the tasks.Originality/valueWork is original.

AbstractMulti-output Gaussian processes (MOGPs) can help to improve predictive performance for some output variables, by leveraging the correlation with other output variables. In this paper, our main motivation is to use multiple-output Gaussian processes to exploit correlations between outputs where each output is a multi-class classification problem. MOGPs have been mostly used for multi-output regression. There are some existing works that use MOGPs for other types of outputs, e.g., multi-output binary classification. However, MOGPs for multi-class classification has been less studied. The reason is twofold: 1) when using a softmax function, it is not clear how to scale it beyond the case of a few outputs; 2) most common type of data in multi-class classification problems consists of image data, and MOGPs are not specifically designed to image data. We thus propose a new MOGPs model called Multi-output Gaussian Processes with Augment & Reduce (MOGPs-AR) that can deal with large scale classification and downsized image input data. Large scale classification is achieved by subsampling both training data sets and classes in each output whereas downsized image input data is handled by incorporating a convolutional kernel into the new model. We show empirically that our proposed model outperforms single-output Gaussian processes in terms of different performance metrics and multi-output Gaussian processes in terms of scalability, both in synthetic and in real classification problems. We include an example with the Ommiglot dataset where we showcase the properties of our model.

Abstract We investigate joint modelling of longevity trends using the spatial statistical framework of Gaussian process (GP) regression. Our analysis is motivated by the Human Mortality Database (HMD) that provides unified raw mortality tables for nearly 40 countries. Yet few stochastic models exist for handling more than two populations at a time. To bridge this gap, we leverage a spatial covariance framework from machine learning that treats populations as distinct levels of a factor covariate, explicitly capturing the cross-population dependence. The proposed multi-output GP models straightforwardly scale up to a dozen populations and moreover intrinsically generate coherent joint longevity scenarios. In our numerous case studies, we investigate predictive gains from aggregating mortality experience across nations and genders, including by borrowing the most recently available “foreign” data. We show that in our approach, information fusion leads to more precise (and statistically more credible) forecasts. We implement our models in R, as well as a Bayesian version in Stan that provides further uncertainty quantification regarding the estimated mortality covariance structure. All examples utilise public HMD datasets.

Multi-output regression seeks to borrow strength and leverage commonalities across different but related outputs in order to enhance learning and prediction accuracy. A fundamental assumption is that the output/group membership labels for all observations are known. This assumption is often violated in real applications. For instance, in healthcare data sets, sensitive attributes such as ethnicity are often missing or unreported. To this end, we introduce a weakly supervised multi-output model based on dependent Gaussian processes. Our approach is able to leverage data without complete group labels or possibly only prior belief on group memberships to enhance accuracy across all outputs. Through intensive simulations and case studies on insulin, testosterone and body fat data sets, we show that our model excels in multi-output settings with missing labels while being competitive in traditional fully labeled settings. We end by highlighting the possible use of our approach in fair inference and sequential decision making.

AbstractGaussian processes occupy one of the leading places in modern statistics and probability theory due to their importance and a wealth of strong results. The common use of Gaussian processes is in connection with problems related to estimation, detection, and many statistical or machine learning models. In this paper, we propose a precise definition of multivariate Gaussian processes based on Gaussian measures on vector-valued function spaces, and provide an existence proof. In addition, several fundamental properties of multivariate Gaussian processes, such as stationarity and independence, are introduced. We further derive two special cases of multivariate Gaussian processes, including multivariate Gaussian white noise and multivariate Brownian motion, and present a brief introduction to multivariate Gaussian process regression as a useful statistical learning method for multi-output prediction problems.