To see the other types of publications on this topic, follow the link: Data distribution shift.
Journal articles on the topic 'Data distribution shift'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 50 journal articles for your research on the topic 'Data distribution shift.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse journal articles on a wide variety of disciplines and organise your bibliography correctly.
1
Cheng, Luling, Xue Yang, Luliang Tang, Qian Duan, Zihan Kan, Xia Zhang, and Xinyue Ye. "Spatiotemporal Analysis of Taxi-Driver Shifts Using Big Trace Data." ISPRS International Journal of Geo-Information 9, no. 4 (April 24, 2020): 281. http://dx.doi.org/10.3390/ijgi9040281.
Full textAbstract:
In taxi management, taxi-driver shift behaviors play a key role in arranging the operation of taxis, which affect the balance between the demand and supply of taxis and the parking spaces. At the same time, these behaviors influence the daily travel of citizens. An analysis of the distribution of taxi-driver shifts, therefore, contributes to transportation management. Compared to the previous research using the real shift records, this study focuses on the spatiotemporal analysis of taxi-driver shifts using big trace data. A two-step strategy is proposed to automatically identify taxi-driver shifts from big trace data without the information of drivers’ identities. The first step is to pick out the frequent spatiotemporal sequential patterns from all parking events based on the spatiotemporal sequence analysis. The second step is to construct a Gaussian mixture model based on prior knowledge for further identifying taxi-driver shifts from all frequent spatiotemporal sequential patterns. The spatiotemporal distribution of taxi-driver shifts is analyzed based on two indicators, namely regional taxi coverage intensity and taxi density. Taking the city of Wuhan as an example, the experimental results show that the identification precision and recall rate of taxi-driver shift events based on the proposed method can achieve about 95% and 90%, respectively, by using big taxi trace data. The occurrence time of taxi-driver shifts in Wuhan mainly has two high peak periods: 1:00 a.m. to 4:00 a.m. and 4:00 p.m. to 5:00 p.m. Although taxi-driver shift behaviors are prohibited during the evening peak hour based on the regulation issued by Wuhan traffic administration, experimental results show that there are still some drivers in violation of this regulation. By analyzing the spatial distribution of taxi-driver shifts, we find that most taxi-driver shifts distribute in central urban areas such as Wuchang and Jianghan district.
2
Islind, Anna Sigridur, Tomas Lindroth, Johan Lundin, and Gunnar Steineck. "Shift in translations: Data work with patient-generated health data in clinical practice." Health Informatics Journal 25, no. 3 (March 13, 2019): 577–86. http://dx.doi.org/10.1177/1460458219833097.
Full textAbstract:
This article reports on how the introduction of patient-generated health data affects the nurses’ and patients’ data work and unpacks how new forms of data collection trigger shifts in the work with data through translation work. The article is based on a 2.5-year case study examining data work of nurses and patients at a cancer rehabilitation clinic at a Swedish Hospital in which patient-generated health data are gathered by patients and then used outside and within clinical practice for decision-making. The article reports on how data are prepared and translated, that is, made useful by the nurses and patients. Using patient-generated health data alters the data work and how the translation of data is performed. The shift in work has three components: (1) a shift in question tactics, (2) a shift in decision-making, and (3) a shift in distribution. The data become mobile, and the data work becomes distributed when using patient-generated health data as an active part of care.
3
Sharet, Nir, and Ilan Shimshoni. "Analyzing Data Changes using Mean Shift Clustering." International Journal of Pattern Recognition and Artificial Intelligence 30, no. 07 (May 25, 2016): 1650016. http://dx.doi.org/10.1142/s0218001416500166.
Full textAbstract:
A nonparametric unsupervised method for analyzing changes in complex datasets is proposed. It is based on the mean shift clustering algorithm. Mean shift is used to cluster the old and new datasets and compare the results in a nonparametric manner. Each point from the new dataset naturally belongs to a cluster of points from its dataset. The method is also able to find to which cluster the point belongs in the old dataset and use this information to report qualitative differences between that dataset and the new one. Changes in local cluster distribution are also reported. The report can then be used to try to understand the underlying reasons which caused the changes in the distributions. On the basis of this method, a transductive transfer learning method for automatically labeling data from the new dataset is also proposed. This labeled data is used, in addition to the old training set, to train a classifier better suited to the new dataset. The algorithm has been implemented and tested on simulated and real (a stereo image pair) datasets. Its performance was also compared with several state-of-the-art methods.
4
Tyagi, Dushyant. "Designing an Effective Combined Shewhart-CUSUM Control Scheme with Exponentially Distributed Data." International Journal of Mathematical, Engineering and Management Sciences 4, no. 5 (October 1, 2019): 1277–86. http://dx.doi.org/10.33889/ijmems.2019.4.5-101.
Full textAbstract:
In this paper, the Combined Shewhart-CUSUM control scheme has been proposed to monitor the production process when the quality characteristic follows exponential distribution to quickly detect the shift in the process. The simulated values of ARL are determined after the transformation of the data into approximate normal distribution by Nelson transformation method and adding Shewhart control limits to existing CUSUM Control Chart. Scheme parameters (value of k and h) and out of control ARL are calculated at various shift and at various in-control ARL. Parameters are also calculated to detect δ standard deviation shifts, which may be helpful to the quality control practitioners in designing the Combined Shewhart-CUSUM scheme when data is highly skewed.
5
Kuang, Kun, Hengtao Zhang, Runze Wu, Fei Wu, Yueting Zhuang, and Aijun Zhang. "Balance-Subsampled Stable Prediction Across Unknown Test Data." ACM Transactions on Knowledge Discovery from Data 16, no. 3 (June 30, 2022): 1–21. http://dx.doi.org/10.1145/3477052.
Full textAbstract:
In data mining and machine learning, it is commonly assumed that training and test data share the same population distribution. However, this assumption is often violated in practice because of the sample selection bias, which might induce the distribution shift from training data to test data. Such a model-agnostic distribution shift usually leads to prediction instability across unknown test data. This article proposes a novel balance-subsampled stable prediction (BSSP) algorithm based on the theory of fractional factorial design. It isolates the clear effect of each predictor from the confounding variables. A design-theoretic analysis shows that the proposed method can reduce the confounding effects among predictors induced by the distribution shift, improving both the accuracy of parameter estimation and the stability of prediction across unknown test data. Numerical experiments on synthetic and real-world datasets demonstrate that our BSSP algorithm can significantly outperform the baseline methods for stable prediction across unknown test data.
6
Ye, Nanyang, Lin Zhu, Jia Wang, Zhaoyu Zeng, Jiayao Shao, Chensheng Peng, Bikang Pan, Kaican Li, and Jun Zhu. "Certifiable Out-of-Distribution Generalization." Proceedings of the AAAI Conference on Artificial Intelligence 37, no. 9 (June 26, 2023): 10927–35. http://dx.doi.org/10.1609/aaai.v37i9.26295.
Full textAbstract:
Machine learning methods suffer from test-time performance degeneration when faced with out-of-distribution (OoD) data whose distribution is not necessarily the same as training data distribution. Although a plethora of algorithms have been proposed to mitigate this issue, it has been demonstrated that achieving better performance than ERM simultaneously on different types of distributional shift datasets is challenging for existing approaches. Besides, it is unknown how and to what extent these methods work on any OoD datum without theoretical guarantees. In this paper, we propose a certifiable out-of-distribution generalization method that provides provable OoD generalization performance guarantees via a functional optimization framework leveraging random distributions and max-margin learning for each input datum. With this approach, the proposed algorithmic scheme can provide certified accuracy for each input datum's prediction on the semantic space and achieves better performance simultaneously on OoD datasets dominated by correlation shifts or diversity shifts. Our code is available at https://github.com/ZlatanWilliams/StochasticDisturbanceLearning.
7
Rezaei, Ashkan, Anqi Liu, Omid Memarrast, and Brian D. Ziebart. "Robust Fairness Under Covariate Shift." Proceedings of the AAAI Conference on Artificial Intelligence 35, no. 11 (May 18, 2021): 9419–27. http://dx.doi.org/10.1609/aaai.v35i11.17135.
Full textAbstract:
Making predictions that are fair with regard to protected attributes (race, gender, age, etc.) has become an important requirement for classification algorithms. Existing techniques derive a fair model from sampled labeled data relying on the assumption that training and testing data are identically and independently drawn (iid) from the same distribution. In practice, distribution shift can and does occur between training and testing datasets as the characteristics of individuals interacting with the machine learning system change. We investigate fairness under covariate shift, a relaxation of the iid assumption in which the inputs or covariates change while the conditional label distribution remains the same. We seek fair decisions under these assumptions on target data with unknown labels. We propose an approach that obtains the predictor that is robust to the worst-case testing performance while satisfying target fairness requirements and matching statistical properties of the source data. We demonstrate the benefits of our approach on benchmark prediction tasks.
8
Soultan, Alaaeldin, Diego Pavón-Jordán, Ute Bradter, Brett K. Sandercock, Wesley M. Hochachka, Alison Johnston, Jon Brommer, et al. "The future distribution of wetland birds breeding in Europe validated against observed changes in distribution." Environmental Research Letters 17, no. 2 (February 1, 2022): 024025. http://dx.doi.org/10.1088/1748-9326/ac4ebe.
Full textAbstract:
Abstract Wetland bird species have been declining in population size worldwide as climate warming and land-use change affect their suitable habitats. We used species distribution models (SDMs) to predict changes in range dynamics for 64 non-passerine wetland birds breeding in Europe, including range size, position of centroid, and margins. We fitted the SDMs with data collected for the first European Breeding Bird Atlas and climate and land-use data to predict distributional changes over a century (the 1970s–2070s). The predicted annual changes were then compared to observed annual changes in range size and range centroid over a time period of 30 years using data from the second European Breeding Bird Atlas. Our models successfully predicted ca. 75% of the 64 bird species to contract their breeding range in the future, while the remaining species (mostly southerly breeding species) were predicted to expand their breeding ranges northward. The northern margins of southerly species and southern margins of northerly species, both, predicted to shift northward. Predicted changes in range size and shifts in range centroids were broadly positively associated with the observed changes, although some species deviated markedly from the predictions. The predicted average shift in core distributions was ca. 5 km yr−1 towards the north (5% northeast, 45% north, and 40% northwest), compared to a slower observed average shift of ca. 3.9 km yr−1. Predicted changes in range centroids were generally larger than observed changes, which suggests that bird distribution changes may lag behind environmental changes leading to ‘climate debt’. We suggest that predictions of SDMs should be viewed as qualitative rather than quantitative outcomes, indicating that care should be taken concerning single species. Still, our results highlight the urgent need for management actions such as wetland creation and restoration to improve wetland birds’ resilience to the expected environmental changes in the future.
9
Lone, Showkat Ahmad, Zahid Rasheed, Sadia Anwar, Majid Khan, Syed Masroor Anwar, and Sana Shahab. "Enhanced fault detection models with real-life applications." AIMS Mathematics 8, no. 8 (2023): 19595–636. http://dx.doi.org/10.3934/math.20231000.
Full textAbstract:
<abstract> <p>Nonconforming events are rare in high-quality processes, and the time between events (TBE) may follow a skewed distribution, such as the gamma distribution. This study proposes one- and two-sided triple homogeneously weighted moving average charts for monitoring TBE data modeled by the gamma distribution. These charts are labeled as the THWMA TBE charts. Monte Carlo simulations are performed to approximate the run length distribution of the one- and two-sided THWMA TBE charts. The THWMA TBE charts are compared to competing charts like the DHWMA TBE, HWMA TBE, THWMA TBE, DEWMA TBE, and EWMA TBE charts at a single shift and over a range of shifts. For the single shift comparison, the average run length (ARL) and standard deviation run length (SDRL) measures are used, whereas the extra quadratic loss (EQL), relative average run length (RARL) and performance comparison index (PCI) measures are employed for a range of shifts comparison. The comparison reveals that the THWMA TBE charts outperform the competing charts at a single shift as well as at a certain range of shifts. Finally, two real-life data applications are presented to evaluate the applicability of the THWMA TBE charts in practical situations, one with boring machine failure data and the other with hospital stay time for traumatic brain injury patients.</p> </abstract>
10
Walther, Gian-Reto, Silje Berger, and Martin T. Sykes. "An ecological ‘footprint’ of climate change." Proceedings of the Royal Society B: Biological Sciences 272, no. 1571 (June 28, 2005): 1427–32. http://dx.doi.org/10.1098/rspb.2005.3119.
Full textAbstract:
Recently, there has been increasing evidence of species' range shifts due to changes in climate. Whereas most of these shifts relate ground truth biogeographic data to a general warming trend in regional or global climate data, we here present a reanalysis of both biogeographic and bioclimatic data of equal spatio-temporal resolution, covering a time span of more than 50 years. Our results reveal a coherent and synchronous shift in both species' distribution and climate. They show not only a shift in the northern margin of a species, which is in concert with gradually increasing winter temperatures in the area, they also confirm the simulated species' distribution changes expected from a bioclimatic model under the recent, relatively moderate climate change.
11
Kuang, Kun, Ruoxuan Xiong, Peng Cui, Susan Athey, and Bo Li. "Stable Prediction with Model Misspecification and Agnostic Distribution Shift." Proceedings of the AAAI Conference on Artificial Intelligence 34, no. 04 (April 3, 2020): 4485–92. http://dx.doi.org/10.1609/aaai.v34i04.5876.
Full textAbstract:
For many machine learning algorithms, two main assumptions are required to guarantee performance. One is that the test data are drawn from the same distribution as the training data, and the other is that the model is correctly specified. In real applications, however, we often have little prior knowledge on the test data and on the underlying true model. Under model misspecification, agnostic distribution shift between training and test data leads to inaccuracy of parameter estimation and instability of prediction across unknown test data. To address these problems, we propose a novel Decorrelated Weighting Regression (DWR) algorithm which jointly optimizes a variable decorrelation regularizer and a weighted regression model. The variable decorrelation regularizer estimates a weight for each sample such that variables are decorrelated on the weighted training data. Then, these weights are used in the weighted regression to improve the accuracy of estimation on the effect of each variable, thus help to improve the stability of prediction across unknown test data. Extensive experiments clearly demonstrate that our DWR algorithm can significantly improve the accuracy of parameter estimation and stability of prediction with model misspecification and agnostic distribution shift.
12
Chan, Chin-Cheng. "Modeling phase shift data of phase-detection autofocus by skew-normal distribution." Journal of Electronic Imaging 28, no. 03 (May 3, 2019): 1. http://dx.doi.org/10.1117/1.jei.28.3.033001.
Full text
13
Gwon, Kyungpil, and Joonhyuk Yoo. "Out-of-Distribution (OOD) Detection and Generalization Improved by Augmenting Adversarial Mixup Samples." Electronics 12, no. 6 (March 16, 2023): 1421. http://dx.doi.org/10.3390/electronics12061421.
Full textAbstract:
Deep neural network (DNN) models are usually built based on the i.i.d. (independent and identically distributed), also known as in-distribution (ID), assumption on the training samples and test data. However, when models are deployed in a real-world scenario with some distributional shifts, test data can be out-of-distribution (OOD) and both OOD detection and OOD generalization should be simultaneously addressed to ensure the reliability and safety of applied AI systems. Most existing OOD detectors pursue these two goals separately, and therefore, are sensitive to covariate shift rather than semantic shift. To alleviate this problem, this paper proposes a novel adversarial mixup (AM) training method which simply executes OOD data augmentation to synthesize differently distributed data and designs a new AM loss function to learn how to handle OOD data. The proposed AM generates OOD samples being significantly diverged from the support of training data distribution but not completely disjoint to increase the generalization capability of the OOD detector. In addition, the AM is combined with a distributional-distance-aware OOD detector at inference to detect semantic OOD samples more efficiently while being robust to covariate shift due to data tampering. Experimental evaluation validates that the designed AM is effective on both OOD detection and OOD generalization tasks compared to previous OOD detectors and data mixup methods.
14
Phongsasiri, Siriwan, and Suwanna Rasmequan. "Outlier Detection in Wellness Data using Probabilistic Mapped Mean-Shift Algorithms." ECTI Transactions on Computer and Information Technology (ECTI-CIT) 15, no. 2 (August 11, 2021): 258–66. http://dx.doi.org/10.37936/ecti-cit.2021152.244971.
Full textAbstract:
In this paper, the Probabilistic Mapped Mean-Shift Algorithm is proposed to detect anomalous data in public datasets and local hospital children’s wellness clinic databases. The proposed framework consists of two main parts. First, the Probabilistic Mapping step consists of k-NN instance acquisition, data distribution calculation, and data point reposition. Truncated Gaussian Distribution (TGD) was used for controlling the boundary of the mapped points. Second, the Outlier Detection step consists of outlier score calculation and outlier selection. Experimental results show that the proposed algorithm outperformed the existing algorithms with real-world benchmark datasets and a Children’s Wellness Clinic dataset (CWD). Outlier detection accuracy obtained from the proposed algorithm based on Wellness, Stamps, Arrhythmia, Pima, and Parkinson datasets was 93%, 94%, 80%, 75%, and 72%, respectively.
15
Monich, Yury V., and Yury D. Nechipurenko. "Correlations in Compositional Data without Log Transformations." Axioms 12, no. 12 (November 27, 2023): 1084. http://dx.doi.org/10.3390/axioms12121084.
Full textAbstract:
This article proposes a method for determining the p-value of correlations in compositional data, i.e., those data that arise as a result of dividing original values by their sum. Data organized in this way are typical for many fields of knowledge, but there is still no consensus on methods for interpreting correlations in such data. In the second decade of the new millennium, almost all newly emerging methods for solving this problem have become based on the log transformation of data. In the method proposed here, there are no log transformations. We return to the early stages of attempting to solve the problem and rely on negative shifts in correlations in the multinomial distribution. In modeling the data, we use a hybrid method that combines the hypergeometric distribution with the distribution of any other law. During our work on the calculation method, we found that the number of degrees of freedom in compositional data measures discretely only when all normalizing sums are equal and that it decreases when the sums are not equal, becoming a continuously varying quantity. Estimation of the number of degrees of freedom and the strength of its influence on the magnitude of the shift in the distribution of correlation coefficients is the basis of the proposed method.
16
Guan, Lisha, Yong Chen, Kevin W. Staples, Jie Cao, and Bai Li. "The influence of complex structure on the spatial dynamics of Atlantic cod (Gadus morhua) in the Gulf of Maine." ICES Journal of Marine Science 74, no. 9 (April 27, 2017): 2379–88. http://dx.doi.org/10.1093/icesjms/fsx064.
Full textAbstract:
Abstract Atlantic cod (Gadus morhua) in the Gulf of Maine (GOM) is an iconic marine fishery stock that has experienced a substantial distributional shift since the mid-1990s. A geostatistical delta-generalized linear mixed model was utilized to hindcast yearly season-specific distributions of GOM cod. These distributions were calculated using the spring and fall bottom trawl survey data for the stock, along with cell-based bathymetry and bottom temperature data for the study area for the years 1982–2013. The centre of stock distribution (the centre of gravity), spatial extent in latitude and longitude, area occupied and median habitat temperature were estimated annually to quantify changes in the spatial dynamics of GOM cod. Time series of these distributional metrics were then used to evaluate the influences of climate change and density-dependent habitat selection on GOM cod’s distribution. Results showed that the rapid southwestward shift in the stock distribution after the late 1990s could not simply be attributed to decreasing stock abundance or warming bottom temperatures. The observed shift in cod distribution requires further investigation on whether it is possibly a result of other factors, like fluctuating productivity among subpopulations.
17
Engelhard, Georg H., John K. Pinnegar, Laurence T. Kell, and Adriaan D. Rijnsdorp. "Nine decades of North Sea sole and plaice distribution." ICES Journal of Marine Science 68, no. 6 (April 11, 2011): 1090–104. http://dx.doi.org/10.1093/icesjms/fsr031.
Full textAbstract:
Abstract Engelhard, G. H., Pinnegar, J. K., Kell, L. T., and Rijnsdorp, A. D. 2011. Nine decades of North Sea sole and plaice distribution. – ICES Journal of Marine Science, 68: 1090–1104. Recent studies based mainly on research survey data suggest that within the North Sea, sole Solea solea and plaice Pleuronectes platessa have exhibited distribution shifts in recent decades—on average southward for sole and northward to deeper waters for plaice. Various hypotheses may account for such shifts, including climate change effects and more intensive fishing in southern and shallower waters; but the relatively short time-span of datasets analysed so far (∼3 decades) has complicated the separation of these two effects. We have made use of a unique dataset of catch and effort data for British North Sea trawlers; these cover nine decades (spanning the period 1913–2007) and are spatially detailed by ICES rectangle (0.5° latitude by 1° longitude). We quantify, for the first time, long-term distribution changes of North Sea sole and plaice over a period approaching a century, and demonstrate that the distribution shift in plaice was attributable to climate change rather than to fishing, but that both climate and fishing played a role in the distribution shift of sole. The discussion also highlights the potential impact of additional factors, including eutrophication, prey availability, and habitat modification.
18
Bai, Haoyue, Rui Sun, Lanqing Hong, Fengwei Zhou, Nanyang Ye, Han-Jia Ye, S. H. Gary Chan, and Zhenguo Li. "DecAug: Out-of-Distribution Generalization via Decomposed Feature Representation and Semantic Augmentation." Proceedings of the AAAI Conference on Artificial Intelligence 35, no. 8 (May 18, 2021): 6705–13. http://dx.doi.org/10.1609/aaai.v35i8.16829.
Full textAbstract:
While deep learning demonstrates its strong ability to handle independent and identically distributed (IID) data, it often suffers from out-of-distribution (OoD) generalization, where the test data come from another distribution (w.r.t. the training one). Designing a general OoD generalization framework for a wide range of applications is challenging, mainly due to different kinds of distribution shifts in the real world, such as the shift across domains or the extrapolation of correlation. Most of the previous approaches can only solve one specific distribution shift, leading to unsatisfactory performance when applied to various OoD benchmarks. In this work, we propose DecAug, a novel decomposed feature representation and semantic augmentation approach for OoD generalization. Specifically, DecAug disentangles the category-related and context-related features by orthogonalizing the two gradients (w.r.t. intermediate features) of losses for predicting category and context labels, where category-related features contain causal information of the target object, while context-related features cause distribution shifts between training and test data. Furthermore, we perform gradient-based augmentation on context-related features to improve the robustness of learned representations. Experimental results show that DecAug outperforms other state-of-the-art methods on various OoD datasets, which is among the very few methods that can deal with different types of OoD generalization challenges.
19
Cheung, William W. L., Jessica J. Meeuwig, Ming Feng, Euan Harvey, Vicky W. Y. Lam, Tim Langlois, Dirk Slawinski, Chaojiao Sun, and Daniel Pauly. "Climate-change induced tropicalisation of marine communities in Western Australia." Marine and Freshwater Research 63, no. 5 (2012): 415. http://dx.doi.org/10.1071/mf11205.
Full textAbstract:
A major observed and predicted impact of climate change on marine species is the poleward shift in their distributions and the resulting changes in community structure. Here, we used a Dynamic Bioclimate Envelope Model to project range shift of exploited marine fishes and invertebrates in Western Australia. We combined published data and expert knowledge to predict current species distributions for 30 tropical, sub-tropical and temperate species that occur along the coast of Western Australia. Using outputs from both a Regional Oceanographic Model and a Global Circulation Model, we simulated change in the distribution of each species. Our study shows that under the SRES (Special Report for Emission Scenarios) A1B scenario, the median rate of distribution shift is around 19 km decade–1 towards higher latitudes and 9 m deeper decade–1 by 2055 relative to 2005. As a result, species gains and losses are expected along the south coast and north coast of Western Australia, respectively. Also, the coast of Western Australia is expected to experience a ‘tropicalisation’ of the marine community in the future, with increasing dominance of warmer-water species. Such changes in species assemblages may have large ecological and socio-economic implications through shifts in fishing grounds and unexpected trophic effects.
20
Chang, Jyun-Ping, Jun-Hong Weng, Hsun-Ching Hsu, Pei-Yuan Lee, and Pin Han. "A Data Transmission Method with Spectral Switches via Electroabsorption." Applied Sciences 12, no. 3 (January 18, 2022): 979. http://dx.doi.org/10.3390/app12030979.
Full textAbstract:
In the past, the waveguide electroabsorption effect has generally been used as an intensity modulator for quasi-monochromatic light, such as lasers. Here, we study how this effect affects polychromatic light spectra. We find that for light with a Gaussian distribution spectrum, the spectral peak shift (red shift or blue shift) can be controlled by the magnitude of the applied voltage, as long as the center wavelength and the spectral band are properly selected. This result can be used as a data transmission scheme at the integrated chip level or in free space. It may offer a good option for some other light sources, such as low-cost LED or ELED (edge emitting LED), with wider spectral bandwidths.
21
Vestfals, Cathleen D., Lorenzo Ciannelli, and Gerald R. Hoff. "Changes in habitat utilization of slope-spawning flatfish across a bathymetric gradient." ICES Journal of Marine Science 73, no. 7 (July 1, 2016): 1875–89. http://dx.doi.org/10.1093/icesjms/fsw112.
Full textAbstract:
Abstract Understanding how fish distributions may change in response to environmental variability is important for effective management of fish populations, as predicted climate change will likely alter their habitat use and population dynamics. This research focused on two commercially- and ecologically-important flatfish species in the eastern Bering Sea (EBS), Greenland halibut ( Reinhardtius hippoglossoides ) and Pacific halibut ( Hippoglossus stenolepis ), which may be especially sensitive to climate-induced shifts in habitat due to strong seasonally and ontogenetically variable distributions. We analysed data from fishery-dependent and fishery-independent sources to determine how environmental variability influenced habitat use, thus gaining a uniquely comprehensive range of seasonal and geographic coverage of each species’ distribution. Greenland and Pacific halibut exhibited strong and contrasting responses to changes in temperature on the shelf, with catches decreasing and increasing, respectively, beyond 1 °C. The effect of temperature was not as prominent along the slope, suggesting that slope habitats may provide some insulation from shelf-associated environmental variability, particularly for Greenland halibut. With warming, Greenland halibut exhibited more of a bathymetric shift in distribution, while the shift was more latitudinal for Pacific halibut. Our results suggest that habitat partitioning may, in part, explain differences in Greenland and Pacific halibut distributions. This research adds to our understanding of how the distributions of two fish species at opposite extremes of their ranges in the EBS – Greenland halibut at the southernmost edge and Pacific halibut at the northernmost edge – may shift in relation to a changing ocean environment.
22
Tomov, Mite, Leonard Abazovski, and Anastasija Ignjatovska. "CONTRIBUTION TO THE SPC IMPLEMENTATION BY SELECTING AN APPROPRIATE VARIABLE CONTROL CHART." Journal of Production Engineering 24, no. 1 (June 30, 2021): 50–54. http://dx.doi.org/10.24867/jpe-2021-01-050.
Full textAbstract:
The research in this paper contributes to the practical implementation of the Statistical Process Control (SPC), primarily in the serial production by selecting an appropriate variable control chart. The paper proposes a procedure algorithm where the deciding criteria used to select the variable control chart include: the data distribution type (whether an approximately normal distribution or not), the number of elements in the subgroup (n) for grouped data, the shift size value and the percentage difference between the shift size value and the mean value of the shifts calculated for each subgroup. The paper explains the proposed algorithm using examples with appropriately drawn control charts, which algorithm essentially represents an extension of the algorithm presented in ISO 7870-2:2013.
23
Jeong, Il-Kyoung, M. J. Graf, and R. H. Heffner. "Effects of Bragg peak profiles and nanoparticle sizes on the real-space pair distribution function." Journal of Applied Crystallography 38, no. 1 (January 19, 2005): 55–61. http://dx.doi.org/10.1107/s0021889804025841.
Full textAbstract:
A study of the effects of Bragg peak profiles and nanoparticle size broadening on the real-space pair distribution function (PDF) is presented, using `synthetic' powder diffraction data. Bragg peak profiles from both asymmetric time-of-flight (TOF) spallation neutron data and symmetric synchrotron X-ray data are considered. Due to their asymmetric peak profiles, the TOF data cause artificial shifts of the PDF peak positions towards higher pair distances. Coupled with this effect is a broadening of the PDF peak widths due to aQ-dependent spectrometer resolution, making reliable refinement of thermal parameters difficult. These effects become more pronounced as theQresolution becomes worse. By contrast, the symmetric X-ray powder diffraction data do not cause a systematic shift of the PDF peak positions, and the broadening of the PDF peak widths has a relatively minor effect on the extraction of the thermal parameters. Finally, nanoparticle size broadening of the asymmetric neutron TOF powder diffraction data causes a shift of the PDF peak positions towards lowerrvalues and smears the PDF intensities from one atomic shell to another.
24
Odedina, Joe. "Online Streaming: A Paradigm Shift for Nollywood Movie Distribution." International Journal of Current Research in the Humanities 27, no. 1 (April 30, 2024): 159–73. http://dx.doi.org/10.4314/ijcrh.v27i1.10.
Full textAbstract:
The Nigerian movie distribution and exhibition business has undergone major transformation in the last few years due to technological innovation. The study is about Nollywood paradigm shift from hard copy movie distribution to online streaming services. The aim is to ascertain why producers are moving to soft screening through online pay-to-view platforms. The objectives were to find out if this new trend is financially beneficial to the producers and whether it will likely continue in the future. The study adopted qualitative methods to interrogate existing data on some of the most popular Nollywood films released online by Netflix. This is based on a snap-shot of Netflix’s popularity chart in the month of October, 2020 and news of cinema theatre films release around the same period. The findings are that the online distribution deals must have been profitable for the producers, judging by repeated use by some of them and others enlisting in the business. The expanding scope of Nollywood producers using these online streaming platforms for distributions of their work suggests the likelihood of a continuing process. The study concludes that this is a mutually beneficial relationship between local producers and online business global entities. It, therefore, recommends that Nollywood producers avoid signing exclusive deals with online distributors only. This will allow home-grown cinema theatres to be financially rewarding to prevent the nation’s cinema theatre chains from collapsing due to redundancy.
25
Das, Nandini. "Process control for categorical (ordinal) data." International Journal of Engineering, Science and Technology 14, no. 2 (November 3, 2022): 34–40. http://dx.doi.org/10.4314/ijest.v14i2.4.
Full textAbstract:
Quality improvement is playing the key role in the success of a business. Reduction of variability is the main step for improvement of quality. Control charts are developed for the purpose of monitoring the quality characteristics with the aim of reducing variability. In many industries instead of continuous variable categorical (ordinal) data are used to measure the quality characteristics of interest. Hence developing control charts techniques for monitoring ordinal data has become a recent research focus. Quality control practitioners often face a problem to select the appropriate technique for monitoring ordinal data in the practical field since there are quite a few techniques available in the literature for this purpose. In this paper we have studied the various techniques for monitoring ordinal data and compared their performance to detect the shift in location parameter. Data were simulated from Normal distribution and average run length (ARL) were computed for different values of shift in mean (both in positive and negative direction) using different methodologies under study. The best technique to detect the shift was identified with respect to ARL.
26
Li, Bao-Chun, and Ya-Zhou Wang. "Charged-Hadron Pseudorapidity Distributions in p-p and Pb-Pb Collisions at LHC Energies." Advances in High Energy Physics 2013 (2013): 1–6. http://dx.doi.org/10.1155/2013/515420.
Full textAbstract:
Pseudorapidity distributions of charged hadrons produced in p-p and Pb-Pb collisions at LHC energies were measured by the CMS and ALICE Collaborations, respectively. An improved Tsallis distribution in the two-cylinder model is used to describe the pseudorapidity spectra. We consider the rapidity shift at the longitudinal direction in the geometrical picture of the collisions. The calculated results are in agreement with the experimental data. The gap between the projectile cylinder and the target cylinder increases with the centralities. The rapidity shifts in the cylinders also increase with the centralities.
27
Omoruyi, Frederick A., Ifunanya Lydia Omeje, Ifeanyi Charles Anabike, and Okechukwu J. Obulezi. "A New Variant of Rama Distribution with Simulation Study and Application to Blood Cancer Data." European Journal of Theoretical and Applied Sciences 1, no. 4 (July 6, 2023): 389–409. http://dx.doi.org/10.59324/ejtas.2023.1(4).36.
Full textAbstract:
In this paper, we propose a new lifetime distribution with flexibility in modeling than its parent distribution. The new distribution is a variant of the Rama distribution having a positive shift parameter. We call the proposed distribution Shifted Rama (SR) distribution. Mathematical and statistical characteristics such as crude moments, central moment, coefficient of variation, index of dispersion, conditional moment, mean residual life function, mean deviation, Bonferroni and Lorenz curve, and the order statistics are derived. Furthermore, reliability measures like survival function, hazard function have been derived. Estimation techniques namely; the maximum likelihood, least squares, weighted least squares, maximum product spacing, Cramer-von- Mises, Anderson-Darling and the right-tailed Anderson-Darling estimations are used. To demonstrate the applicability of the distribution, a numerical example was the blood cancer data from Ministry Hospital in Saudi Arabia. Based on the results, the proposed distribution performed better than the competing distributions. Simulation of the Estimates of the parameters based on the classical methods considered are obtained, and result showed that the maximum likelihood estimator gave the best classical estimates of the parameters compared to other methods considered.
28
Martínez-Martínez, Josué, Olivia Brown, and Rajmonda Caceres. "Towards Robustness to Natural Variations and Distribution Shift (Student Abstract)." Proceedings of the AAAI Conference on Artificial Intelligence 38, no. 21 (March 24, 2024): 23579–81. http://dx.doi.org/10.1609/aaai.v38i21.30481.
Full textAbstract:
This research focuses on improving the robustness of machine learning systems to natural variations and distribution shifts. A design trade space is presented, and various methods are compared, including adversarial training, data augmentation techniques, and novel approaches inspired by model-based robust optimization formulations.
29
Alexander, Chinago B., and Vincent E. Weli. "Statistical Analyses of Rainfall Distribution: A Check on Climatic Shift." Journal of Progress in Engineering and Physical Science 2, no. 1 (March 2023): 1–12. http://dx.doi.org/10.56397/jpeps.2023.03.01.
Full textAbstract:
The study Statistical Analyse of Rainfall Distribution: A check on Climate Shift is concerned with using table values and graph to visually compare rainfall distribution over different climatic cycles. To achieve this, rainfall data from Nigeria Meteorological Agency archive was extracted for analysis. The study observed that only two months had mean rainfall less than 51mm. It was also observed that rainfall occurrence decreases over the years, as indicated by the negative trend line. The study among other things discovered that the double maxima as a result of August dry spell has reduced significantly, thereby increasing run-off. This paper observed that the time of the years is significantly related to rainfall occurrence. The month of August had more improved rainfall throughout the period of study. The rate of dispersion of mean annual rainfall from the station mean is an indication of the reliability of rainfall or otherwise. This work opined that change in climate will lead to a complete shift in RF distributions; therefore, the activities that were influenced by RF will be affected positively or in a negative way in the future. This work recommends that agriculture calendar of the study area need to be changed therefore farmers need to adopt new farming techniques as RF distribution change. There agricultural calendar needs to also change.
30
Culpeper, Jonathan, and Merja Kytö. "Data in historical pragmatics." Journal of Historical Pragmatics 1, no. 2 (August 30, 2000): 175–99. http://dx.doi.org/10.1075/jhp.1.2.03cul.
Full textAbstract:
In this paper we examine four speech-related text types in terms of how linguistically close they are to spoken face-to-face interaction. Our “conversational” diagnostics include lexical repetitions, question marks (as an indicator of question-answer adjacency pairs), interruptions, and several single word interactive features (first- and second-person pronouns, private verbs and demonstrative pronouns). We discuss the nature of these diagnostics and then consider their distribution across our text types and across the period 1600 to 1720. We reveal: (1) a differential distribution across our text types (and suggest a number of explanatory factors), and (2) a shift over our period towards features associated with spoken face-to-face interaction (and make the tentative suggestion that this finding may be due to the development of “popular” literatures). We also make some preliminary remarks about our Shakespeare sample.
31
Zhou, Fang Bin, and Yun Kai Guo. "Analysis on Difference of Contaminated Normal Distribution PDF." Applied Mechanics and Materials 409-410 (September 2013): 1661–66. http://dx.doi.org/10.4028/www.scientific.net/amm.409-410.1661.
Full textAbstract:
As a very important distribution, contaminated normal distribution play a great role in data processing. The probability density function (PDF) feature of the contaminated normal distribution was investigated. The Kullback-Leibler distance is suggested for measuring PDF difference between mean shift model and variance inflation model. Numerical calculations show that the PDF difference of two kinds of model is related to mean shift parameter λ and the variance inflation factor α closely when the main distribution is the standard normal distribution and the relationship is nonlinear proportional.
32
van der Kooij, Jeroen, Sascha M. M. Fässler, David Stephens, Lisa Readdy, Beth E. Scott, and Beatriz A. Roel. "Opportunistically recorded acoustic data support Northeast Atlantic mackerel expansion theory." ICES Journal of Marine Science 73, no. 4 (December 19, 2015): 1115–26. http://dx.doi.org/10.1093/icesjms/fsv243.
Full textAbstract:
Abstract Fisheries independent monitoring of widely distributed pelagic fish species which conduct large seasonal migrations is logistically complex and expensive. One of the commercially most important examples of such a species in the Northeast Atlantic Ocean is mackerel for which up to recently only an international triennial egg survey contributed to the stock assessment. In this study, we explore whether fisheries acoustic data, recorded opportunistically during the English component of the North Sea International Bottom Trawl Survey, can contribute to an improved understanding of mackerel distribution and provide supplementary data to existing dedicated monitoring surveys. Using a previously published multifrequency acoustic mackerel detection algorithm, we extracted the distribution and abundance of schooling mackerel for the whole of the North Sea during August and September between 2007 and 2013. The spatio-temporal coverage of this unique dataset is of particular interest because it includes part of the unsurveyed summer mackerel feeding grounds in the northern North Sea. Recent increases in landings in Icelandic waters during this season suggested that changes have occurred in the mackerel feeding distribution. Thus far it is poorly understood whether these changes are due to a shift, i.e. mackerel moving away from their traditional feeding grounds in the northern North Sea and southern Norwegian Sea, or whether the species' distribution has expanded. We therefore explored whether acoustically derived biomass of schooling mackerel declined in the northern North Sea during the study period, which would suggest a shift in mackerel distribution rather than an expansion. The results of this study show that in the North Sea, schooling mackerel abundance has increased and that its distribution in this area has not changed over this period. Both of these findings provide, to our knowledge, the first evidence in support of the hypothesis that mackerel have expanded their distribution rather than moved away.
33
Weinman, J. A., F. S. Marzano, W. J. Plant, A. Mugnai, and N. Pierdicca. "Rainfall observation from X-band, space-borne, synthetic aperture radar." Natural Hazards and Earth System Sciences 9, no. 1 (February 4, 2009): 77–84. http://dx.doi.org/10.5194/nhess-9-77-2009.
Full textAbstract:
Abstract. Satellites carrying X-band Synthetic Aperture Radars (SAR) have recently been launched by several countries. These provide new opportunities to measure precipitation with higher spatial resolution than has heretofore been possible. Two algorithms to retrieve precipitation from such measurements over land have been developed, and the retrieved rainfall distributions were found to be consistent. A maritime rainfall distribution obtained from dual frequency (X and C-band) data was used to compute the Differential Polarized Phase Shift. The computed Differential Polarized Phase Shift compared well with the value measured from space. Finally, we show a comparison between a recent X-band SAR image of a precipitation distribution and an observation of the same rainfall from ground-based operational weather radar. Although no quantitative comparison of retrieved and conventional rainfall distributions could be made with the available data at this time, the results presented here point the way to such comparisons.
34
Deng, Junli, Haoyuan Yao, and Ping Shi. "Enhanced 3D Pose Estimation in Multi-Person, Multi-View Scenarios through Unsupervised Domain Adaptation with Dropout Discriminator." Sensors 23, no. 20 (October 12, 2023): 8406. http://dx.doi.org/10.3390/s23208406.
Full textAbstract:
Data-driven pose estimation methods often assume equal distributions between training and test data. However, in reality, this assumption does not always hold true, leading to significant performance degradation due to distribution mismatches. In this study, our objective is to enhance the cross-domain robustness of multi-view, multi-person 3D pose estimation. We tackle the domain shift challenge through three key approaches: (1) A domain adaptation component is introduced to improve estimation accuracy for specific target domains. (2) By incorporating a dropout mechanism, we train a more reliable model tailored to the target domain. (3) Transferable Parameter Learning is employed to retain crucial parameters for learning domain-invariant data. The foundation for these approaches lies in the H-divergence theory and the lottery ticket hypothesis, which are realized through adversarial training by learning domain classifiers. Our proposed methodology is evaluated using three datasets: Panoptic, Shelf, and Campus, allowing us to assess its efficacy in addressing domain shifts in multi-view, multi-person pose estimation. Both qualitative and quantitative experiments demonstrate that our algorithm performs well in two different domain shift scenarios.
35
Yan, Yuguang, Mingkui Tan, Yanwu Xu, Jiezhang Cao, Michael Ng, Huaqing Min, and Qingyao Wu. "Oversampling for Imbalanced Data via Optimal Transport." Proceedings of the AAAI Conference on Artificial Intelligence 33 (July 17, 2019): 5605–12. http://dx.doi.org/10.1609/aaai.v33i01.33015605.
Full textAbstract:
The issue of data imbalance occurs in many real-world applications especially in medical diagnosis, where normal cases are usually much more than the abnormal cases. To alleviate this issue, one of the most important approaches is the oversampling method, which seeks to synthesize minority class samples to balance the numbers of different classes. However, existing methods barely consider global geometric information involved in the distribution of minority class samples, and thus may incur distribution mismatching between real and synthetic samples. In this paper, relying on optimal transport (Villani 2008), we propose an oversampling method by exploiting global geometric information of data to make synthetic samples follow a similar distribution to that of minority class samples. Moreover, we introduce a novel regularization based on synthetic samples and shift the distribution of minority class samples according to loss information. Experiments on toy and real-world data sets demonstrate the efficacy of our proposed method in terms of multiple metrics.
36
Vargas, David A., Juan F. De Villena, Valeria Larios, Rossy Bueno López, Daniela R. Chávez-Velado, Diego E. Casas, Reagan L. Jiménez, Sabrina E. Blandon, and Marcos X. Sanchez-Plata. "Data-Mining Poultry Processing Bio-Mapping Counts of Pathogens and Indicator Organisms for Food Safety Management Decision Making." Foods 12, no. 4 (February 20, 2023): 898. http://dx.doi.org/10.3390/foods12040898.
Full textAbstract:
Bio-mapping studies play an important role, as the data collected can be managed and analyzed in multiple ways to look at process trends, find explanations about the effect of process changes, activate a root cause analysis for events, and even compile performance data to demonstrate to inspection authorities or auditors the effect of certain decisions made on a daily basis and their effects over time in commercial settings not only from the food safety perspective but also from the production side. This study presents an alternative analysis of bio-mapping data collected throughout several months in a commercial poultry processing operation as described in the article “Bio-Mapping Indicators and Pathogen Loads in a Commercial Broiler Processing Facility Operating with High and Low Antimicrobial Interventions”. The conducted analysis identifies the processing shift effect on microbial loads, attempts to find correlation between microbial indicators data and pathogens loads, and identifies novel visualization approaches and conducts distribution analysis for microbial indicators and pathogens in a commercial poultry processing facility. From the data analyzed, a greater number of locations were statistically different between shifts under reduced levels of chemical interventions with higher means at the second shift for both indicators and pathogens levels. Minimal to negligible correlation was found when comparing aerobic counts and Enterobacteriaceae counts with Salmonella levels, with significant variability between sampling locations. Distribution analysis and visualization as a bio-map of the process resulted in a clear bimodality in reduced chemical conditions for multiple locations mostly explained by shift effect. The development and use of bio-mapping data, including proper data visualization, improves the tools needed for ongoing decision making in food safety systems.
37
Zhao, Xingyu, Yuexuan An, Ning Xu, Jing Wang, and Xin Geng. "Imbalanced Label Distribution Learning." Proceedings of the AAAI Conference on Artificial Intelligence 37, no. 9 (June 26, 2023): 11336–44. http://dx.doi.org/10.1609/aaai.v37i9.26341.
Full textAbstract:
Label distribution covers a certain number of labels, representing the degree to which each label describes an instance. The learning process on the instances labeled by label distributions is called Label Distribution Learning (LDL). Although LDL has been applied successfully to many practical applications, one problem with existing LDL methods is that they are limited to data with balanced label information. However, annotation information in real-world data often exhibits imbalanced distributions, which significantly degrades the performance of existing methods. In this paper, we investigate the Imbalanced Label Distribution Learning (ILDL) problem. To handle this challenging problem, we delve into the characteristics of ILDL and empirically find that the representation distribution shift is the underlying reason for the performance degradation of existing methods. Inspired by this finding, we present a novel method named Representation Distribution Alignment (RDA). RDA aligns the distributions of feature representations and label representations to alleviate the impact of the distribution gap between the training set and the test set caused by the imbalance issue. Extensive experiments verify the superior performance of RDA. Our work fills the gap in benchmarks and techniques for practical ILDL problems.
38
Jin, Kun, Tongxin Yin, Zhongzhu Chen, Zeyu Sun, Xueru Zhang, Yang Liu, and Mingyan Liu. "Performative Federated Learning: A Solution to Model-Dependent and Heterogeneous Distribution Shifts." Proceedings of the AAAI Conference on Artificial Intelligence 38, no. 11 (March 24, 2024): 12938–46. http://dx.doi.org/10.1609/aaai.v38i11.29191.
Full textAbstract:
We consider a federated learning (FL) system consisting of multiple clients and a server, where the clients aim to collaboratively learn a common decision model from their distributed data. Unlike the conventional FL framework that assumes the client's data is static, we consider scenarios where the clients' data distributions may be reshaped by the deployed decision model. In this work, we leverage the idea of distribution shift mappings in performative prediction to formalize this model-dependent data distribution shift and propose a performative FL framework. We first introduce necessary and sufficient conditions for the existence of a unique performative stable solution and characterize its distance to the performative optimal solution. Then we propose the performative FedAvg algorithm and show that it converges to the performative stable solution at a rate of O(1/T) under both full and partial participation schemes. In particular, we use novel proof techniques and show how the clients' heterogeneity influences the convergence. Numerical results validate our analysis and provide valuable insights into real-world applications.
39
Doskocz, Adam. "Accuracy assessment of planimetric large-scale map data for decision-making." Geodesy and Cartography 65, no. 1 (June 1, 2016): 3–12. http://dx.doi.org/10.1515/geocart-2016-0006.
Full textAbstract:
Abstract This paper presents decision-making risk estimation based on planimetric large-scale map data, which are data sets or databases which are useful for creating planimetric maps on scales of 1:5,000 or larger. The studies were conducted on four data sets of large-scale map data. Errors of map data were used for a risk assessment of decision-making about the localization of objects, e.g. for land-use planning in realization of investments. An analysis was performed for a large statistical sample set of shift vectors of control points, which were identified with the position errors of these points (errors of map data). In this paper, empirical cumulative distribution function models for decision-making risk assessment were established. The established models of the empirical cumulative distribution functions of shift vectors of control points involve polynomial equations. An evaluation of the compatibility degree of the polynomial with empirical data was stated by the convergence coefficient and by the indicator of the mean relative compatibility of model. The application of an empirical cumulative distribution function allows an estimation of the probability of the occurrence of position errors of points in a database. The estimated decision-making risk assessment is represented by the probability of the errors of points stored in the database.
40
Archis, Jennifer N., Christopher Akcali, Bryan L. Stuart, David Kikuchi, and Amanda J. Chunco. "Is the future already here? The impact of climate change on the distribution of the eastern coral snake (Micrurus fulvius)." PeerJ 6 (May 1, 2018): e4647. http://dx.doi.org/10.7717/peerj.4647.
Full textAbstract:
Anthropogenic climate change is a significant global driver of species distribution change. Although many species have undergone range expansion at their poleward limits, data on several taxonomic groups are still lacking. A common method for studying range shifts is using species distribution models to evaluate current, and predict future, distributions. Notably, many sources of ‘current’ climate data used in species distribution modeling use the years 1950–2000 to calculate climatic averages. However, this does not account for recent (post 2000) climate change. This study examines the influence of climate change on the eastern coral snake (Micrurus fulvius). Specifically, we: (1) identified the current range and suitable environment of M. fulvius in the Southeastern United States, (2) investigated the potential impacts of climate change on the distribution of M. fulvius, and (3) evaluated the utility of future models in predicting recent (2001–2015) records. We used the species distribution modeling program Maxent and compared both current (1950–2000) and future (2050) climate conditions. Future climate models showed a shift in the distribution of suitable habitat across a significant portion of the range; however, results also suggest that much of the Southeastern United States will be outside the range of current conditions, suggesting that there may be no-analog environments in the future. Most strikingly, future models were more effective than the current models at predicting recent records, suggesting that range shifts may already be occurring. These results have implications for both M. fulvius and its Batesian mimics. More broadly, we recommend future Maxent studies consider using future climate data along with current data to better estimate the current distribution.
41
An, Shengwei, Sheng-Yen Chou, Kaiyuan Zhang, Qiuling Xu, Guanhong Tao, Guangyu Shen, Siyuan Cheng, et al. "Elijah: Eliminating Backdoors Injected in Diffusion Models via Distribution Shift." Proceedings of the AAAI Conference on Artificial Intelligence 38, no. 10 (March 24, 2024): 10847–55. http://dx.doi.org/10.1609/aaai.v38i10.28958.
Full textAbstract:
Diffusion models (DM) have become state-of-the-art generative models because of their capability of generating high-quality images from noises without adversarial training. However, they are vulnerable to backdoor attacks as reported by recent studies. When a data input (e.g., some Gaussian noise) is stamped with a trigger (e.g., a white patch), the backdoored model always generates the target image (e.g., an improper photo). However, effective defense strategies to mitigate backdoors from DMs are underexplored. To bridge this gap, we propose the first backdoor detection and removal framework for DMs. We evaluate our framework Elijah on over hundreds of DMs of 3 types including DDPM, NCSN and LDM, with 13 samplers against 3 existing backdoor attacks. Extensive experiments show that our approach can have close to 100% detection accuracy and reduce the backdoor effects to close to zero without significantly sacrificing the model utility.
42
Tasche, Dirk. "Factorizable Joint Shift in Multinomial Classification." Machine Learning and Knowledge Extraction 4, no. 3 (September 10, 2022): 779–802. http://dx.doi.org/10.3390/make4030038.
Full textAbstract:
Factorizable joint shift (FJS) was recently proposed as a type of dataset shift for which the complete characteristics can be estimated from feature data observations on the test dataset by a method called Joint Importance Aligning. For the multinomial (multiclass) classification setting, we derive a representation of factorizable joint shift in terms of the source (training) distribution, the target (test) prior class probabilities and the target marginal distribution of the features. On the basis of this result, we propose alternatives to joint importance aligning and, at the same time, point out that factorizable joint shift is not fully identifiable if no class label information on the test dataset is available and no additional assumptions are made. Other results of the paper include correction formulae for the posterior class probabilities both under general dataset shift and factorizable joint shift. In addition, we investigate the consequences of assuming factorizable joint shift for the bias caused by sample selection.
43
Xin, Shiji, Yifei Wang, Jingtong Su, and Yisen Wang. "On the Connection between Invariant Learning and Adversarial Training for Out-of-Distribution Generalization." Proceedings of the AAAI Conference on Artificial Intelligence 37, no. 9 (June 26, 2023): 10519–27. http://dx.doi.org/10.1609/aaai.v37i9.26250.
Full textAbstract:
Despite impressive success in many tasks, deep learning models are shown to rely on spurious features, which will catastrophically fail when generalized to out-of-distribution (OOD) data. Invariant Risk Minimization (IRM) is proposed to alleviate this issue by extracting domain-invariant features for OOD generalization. Nevertheless, recent work shows that IRM is only effective for a certain type of distribution shift (e.g., correlation shift) while it fails for other cases (e.g., diversity shift). Meanwhile, another thread of method, Adversarial Training (AT), has shown better domain transfer performance, suggesting that it has the potential to be an effective candidate for extracting domain-invariant features. This paper investigates this possibility by exploring the similarity between the IRM and AT objectives. Inspired by this connection, we propose Domain-wise Adversarial Training (DAT), an AT-inspired method for alleviating distribution shift by domain-specific perturbations. Extensive experiments show that our proposed DAT can effectively remove domain-varying features and improve OOD generalization under both correlation shift and diversity shift.
44
Cornille, Nathan, Katrien Laenen, Jingyuan Sun, and Marie-Francine Moens. "Causal Factor Disentanglement for Few-Shot Domain Adaptation in Video Prediction." Entropy 25, no. 11 (November 17, 2023): 1554. http://dx.doi.org/10.3390/e25111554.
Full textAbstract:
An important challenge in machine learning is performing with accuracy when few training samples are available from the target distribution. If a large number of training samples from a related distribution are available, transfer learning can be used to improve the performance. This paper investigates how to do transfer learning more effectively if the source and target distributions are related through a Sparse Mechanism Shift for the application of next-frame prediction. We create Sparse Mechanism Shift-TempoRal Intervened Sequences (SMS-TRIS), a benchmark to evaluate transfer learning for next-frame prediction derived from the TRIS datasets. We then propose to exploit the Sparse Mechanism Shift property of the distribution shift by disentangling the model parameters with regard to the true causal mechanisms underlying the data. We use the Causal Identifiability from TempoRal Intervened Sequences (CITRIS) model to achieve this disentanglement via causal representation learning. We show that encouraging disentanglement with the CITRIS extensions can improve performance, but their effectiveness varies depending on the dataset and backbone used. We find that it is effective only when encouraging disentanglement actually succeeds in increasing disentanglement. We also show that an alternative method designed for domain adaptation does not help, indicating the challenging nature of the SMS-TRIS benchmark.
45
Feng, E., Z. Zia, C. Tong, and N. Cornell. "P105: Observational study of distribution of time and activities over the course of an emergency physician's shift." CJEM 22, S1 (May 2020): S102—S103. http://dx.doi.org/10.1017/cem.2020.311.
Full textAbstract:
Introduction: The growing scrutiny to improve Emergency Department (ED) wait times and patient flow have resulted in many efforts to increase efficiency and maximize patient throughput via systems improvements. This study investigates areas of efficiency improvement from the Emergency Physician (EP) perspective by examining EP workflow in a two phased observational time-motion study. In the initial phase, the distribution of time and activities of EPs were dissected to identify potential sources for streamlining to maximize physician productivity. The first phase was of the study was completed during the period immediately preceding the implementation of an Electronic Health Records (EHR). The second phase of the study will repeat the analysis one year post EHR implementation. This data will be dissected to again identify sources for streamlining in an EHR environment and to identify shifts in work flow from a paper-based system. Methods: An observational time motion study was conducted at St. Mary's Hospital ED, in Kitchener Ontario. An observer was paired with an EP for the duration of an 8 hour shift, to a total of 14 shifts in the first phase of the study. Nine task categories were measured concurrently with a stopwatch application on a tablet, along with the number of interruptions experienced by the EP. Means of each category were calculated and converted to percentages, representing the amount of time per 8 hour shift dedicated to each activity. The second phase will be repeated in Fall 2020, 1 year after EHR implementation. Results: A total of 14 shifts were observed, accounting for 112 hours of observation. EP's time was allocated amongst the following categories: direct patient interaction (40.8%), documentation (27.1%), reviewing patient results (18.4%), communicating with ED staff (7.63%), personal activities (5.7%), writing orders (5.1%), communicating with consultants (3.3%), teaching (1.7%) and medical information searches (1.3%). On average, EPs experienced 15.8 interruptions over the course of an 8 hour shift. Conclusion: In a paper charting system, the direct patient interaction accounts for the largest timeshare over the course of a given shift. However, the next two largest categories, documentation and reviewing patient data, both represent areas of potential streamlining via clerical improvements. Additionally, detailed measurements of EPs’ activities have proven feasible and provides the potential for future insight into the impact of EHR's on EP workflow.
46
Bolarinwa, Mojisola A., and Mayowa T. Odetunde. "Appraising the Cognitive Alertness of Night Shift Workers: Case of Bottling Company Operators in Central Northern Nigeria." European Journal of Business and Management Research 8, no. 4 (July 26, 2023): 122–28. http://dx.doi.org/10.24018/ejbmr.2023.8.4.2036.
Full textAbstract:
Night shift, a common work schedule in 24-hour production companies has been posing some after-effects on workers’ health, ranging from physical to psychological, and lack of response to these effects could result in cognitive deficiencies for workers and increased downtime for organisations. The objective of this study is to evaluate the effect of the night shift on the response time of workers to digital visual stimuli as compared to the day shift. Twenty-two shift operators from a bottling line between the ages of 25 and 35 years were subjected to psychomotor vigilance test (PVT) using an android PVT application (Versor-PVT Version 1.03) to assess their response times to visual stimuli and analysed using the SPSS software to determine the data distribution, skewness, and kurtosis. Based on the determined data distribution, Wilcoxon signed rank test (WSRT) was utilised to determine the level of significance of the differences between the two of them. The recorded minimum and maximum average response times for post-day and post-night shifts respectively were: 609.94 ms and 733.64 ms; 644.56 ms and 944.17 ms. Results clearly showed that night shift has significant negative impact on the response time of tested operators, such as delayed ability to respond to situations, including fault rectification, thus causing safety concern, and reducing productivity. In conclusion, additional cost is imminently incurred on the part of the organisation in settling medical bills and compensations. The WSRT analysis indicated that there is 95% probability of generating similar results on repeating the experiment under same conditions.
47
Askelson, Mark A., Patricia M. Pauley, and Jerry M. Straka. "Response Functions for Arbitrary Weight Functions and Data Distributions. Part II: Response Function Derivation and Verification." Monthly Weather Review 133, no. 8 (August 1, 2005): 2132–47. http://dx.doi.org/10.1175/mwr2963.1.
Full textAbstract:
Abstract Distance-dependent weighted averaging (DDWA) is a process that is fundamental to most of the objective analysis schemes that are used in meteorology. Despite its ubiquity, aspects of its effects are still poorly understood. This is especially true for the most typical situation of observations that are discrete, bounded, and irregularly distributed. To facilitate understanding of the effects of DDWA schemes, a framework that enables the determination of response functions for arbitrary weight functions and data distributions is developed. An essential element of this approach is the equivalent analysis, which is a hypothetical analysis that is produced by using, throughout the analysis domain, the same weight function and data distribution that apply at the point where the response function is desired. This artifice enables the derivation of the response function by way of the convolution theorem. Although this approach requires a bit more effort than an alternative one, the reward is additional insight into the impacts of DDWA analyses. An important insight gained through this approach is the exact nature of the DDWA response function. For DDWA schemes the response function is the complex conjugate of the normalized Fourier transform of the effective weight function. In facilitating this result, this approach affords a better understanding of which elements (weight functions, data distributions, normalization factors, etc.) affect response functions and how they interact to do so. Tests of the response function for continuous, bounded data and discrete, irregularly distributed data verify the validity of the response functions obtained herein. They also reinforce previous findings regarding the dependence of response functions on analysis location and the impacts of data boundaries and irregular data spacing. Interpretation of the response function in terms of amplitude and phase modulations is illustrated using examples. Inclusion of phase shift information is important in the evaluation of DDWA schemes when they are applied to situations that may produce significant phase shifts. These situations include those where data boundaries influence the analysis value and where data are irregularly distributed. By illustrating the attendant movement, or shift, of data, phase shift information also provides an elegant interpretation of extrapolation.
48
Schaber, Clemens F., Silja Flenner, Anja Glisovic, Igor Krasnov, Martin Rosenthal, Hergen Stieglitz, Christina Krywka, Manfred Burghammer, Martin Müller, and Stanislav N. Gorb. "Hierarchical architecture of spider attachment setae reconstructed from scanning nanofocus X-ray diffraction data." Journal of The Royal Society Interface 16, no. 150 (January 2019): 20180692. http://dx.doi.org/10.1098/rsif.2018.0692.
Full textAbstract:
When sitting and walking, the feet of wandering spiders reversibly attach to many surfaces without the use of gluey secretions. Responsible for the spiders' dry adhesion are the hairy attachment pads that are built of specially shaped cuticular hairs (setae) equipped with approximately 1 µm wide and 20 nm thick plate-like contact elements (spatulae) facing the substrate. Using synchrotron-based scanning nanofocus X-ray diffraction methods, combining wide-angle X-ray diffraction/scattering and small-angle X-ray scattering, allowed substantial quantitative information to be gained about the structure and materials of these fibrous adhesive structures with 200 nm resolution. The fibre diffraction patterns showed the crystalline chitin chains oriented along the long axis of the attachment setae and increased intensity of the chitin signal dorsally within the seta shaft. The small-angle scattering signals clearly indicated an angular shift by approximately 80° of the microtrich structures that branch off the bulk hair shaft and end as the adhesive contact elements in the tip region of the seta. The results reveal the specific structural arrangement and distribution of the chitin fibres within the attachment hair's cuticle preventing material failure by tensile reinforcement and proper distribution of stresses that arise upon attachment and detachment.
49
Jagielski, Matthew, Stanley Wu, Alina Oprea, Jonathan Ullman, and Roxana Geambasu. "How to Combine Membership-Inference Attacks on Multiple Updated Machine Learning Models." Proceedings on Privacy Enhancing Technologies 2023, no. 3 (July 2023): 211–32. http://dx.doi.org/10.56553/popets-2023-0078.
Full textAbstract:
A large body of research has shown that machine learning models are vulnerable to membership inference (MI) attacks that violate the privacy of the participants in the training data. Most MI research focuses on the case of a single standalone model, while production machine-learning platforms often update models over time, on data that often shifts in distribution, giving the attacker more information. This paper proposes new attacks that take advantage of one or more model updates to improve MI. A key part of our approach is to leverage rich information from standalone MI attacks mounted separately against the original and updated models, and to combine this information in specific ways to improve attack effectiveness. We propose a set of combination functions and tuning methods for each, and present both analytical and quantitative justification for various options. Our results on four public datasets show that our attacks are effective at using update information to give the adversary a significant advantage over attacks on standalone models, but also compared to a prior MI attack that takes advantage of model updates in a related machine-unlearning setting. We perform the first measurements of the impact of distribution shift on MI attacks with model updates, and show that a more drastic distribution shift results in significantly higher MI risk than a gradual shift. We also show that our attacks are effective at auditing differentially private fine tuning. We make our code public on Github: https://github.com/stanleykywu/model-updates.
50
Alferova, E., D. Lychagin, and A. Chernyakov. "Research of Stress Field Distribution in FCC-Single Crystal Samples in Compression." Applied Mechanics and Materials 682 (October 2014): 485–90. http://dx.doi.org/10.4028/www.scientific.net/amm.682.485.
Full textAbstract:
.Theoretical distribution of a stress field in a sample in the form of a rectangular parallelepiped in compression for an isotropic material was calculated using the method of finite elements. Calculations showed that the highest stress is observed at the top and front edges of a sample. There are areas of the tension stress on vertical edges in the area of sample tops. Shift anisotropy was determined by imposing of the distributed tension field in a sample on FCC crystal of a certain crystallographic orientation.Change of shift symmetry in single crystals for different crystallographic orientations of a compression axis was considered. It was established that a shift fragmentation in the parallel octahedral planes in the conditions of plastic deformation determines the process of low-symmetric shift deformation and maintenance of higher single crystal pseudo-symmetry. Connection of the obtained results and test data on heterogeneity of plastic deformation of nickel and aluminum single crystals is discussed.