Dissertations / Theses on the topic 'UCI MACHINE LEARNING'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 50 dissertations / theses for your research on the topic 'UCI MACHINE LEARNING.'
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 dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
Modi, Navikkumar. "Machine Learning and Statistical Decision Making for Green Radio." Thesis, CentraleSupélec, 2017. http://www.theses.fr/2017SUPL0002/document.
Full textFuture cellular network technologies are targeted at delivering self-organizable and ultra-high capacity networks, while reducing their energy consumption. This thesis studies intelligent spectrum and topology management through cognitive radio techniques to improve the capacity density and Quality of Service (QoS) as well as to reduce the cooperation overhead and energy consumption. This thesis investigates how reinforcement learning can be used to improve the performance of a cognitive radio system. In this dissertation, we deal with the problem of opportunistic spectrum access in infrastructureless cognitive networks. We assume that there is no information exchange between users, and they have no knowledge of channel statistics and other user's actions. This particular problem is designed as multi-user restless Markov multi-armed bandit framework, in which multiple users collect a priori unknown reward by selecting a channel. The main contribution of the dissertation is to propose a learning policy for distributed users, that takes into account not only the availability criterion of a band but also a quality metric linked to the interference power from the neighboring cells experienced on the sensed band. We also prove that the policy, named distributed restless QoS-UCB (RQoS-UCB), achieves at most logarithmic order regret. Moreover, numerical studies show that the performance of the cognitive radio system can be significantly enhanced by utilizing proposed learning policies since the cognitive devices are able to identify the appropriate resources more efficiently. This dissertation also introduces a reinforcement learning and transfer learning frameworks to improve the energy efficiency (EE) of the heterogeneous cellular network. Specifically, we formulate and solve an energy efficiency maximization problem pertaining to dynamic base stations (BS) switching operation, which is identified as a combinatorial learning problem, with restless Markov multi-armed bandit framework. Furthermore, a dynamic topology management using the previously defined algorithm, RQoS-UCB, is introduced to intelligently control the working modes of BSs, based on traffic load and capacity in multiple cells. Moreover, to cope with initial reward loss and to speed up the learning process, a transfer RQoS-UCB policy, which benefits from the transferred knowledge observed in historical periods, is proposed and provably converges. Then, proposed dynamic BS switching operation is demonstrated to reduce the number of activated BSs while maintaining an adequate QoS. Extensive numerical simulations demonstrate that the transfer learning significantly reduces the QoS fluctuation during traffic variation, and it also contributes to a performance jump-start and presents significant EE improvement under various practical traffic load profiles. Finally, a proof-of-concept is developed to verify the performance of proposed learning policies on a real radio environment and real measurement database of HF band. Results show that proposed multi-armed bandit learning policies using dual criterion (e.g. availability and quality) optimization for opportunistic spectrum access is not only superior in terms of spectrum utilization but also energy efficient
Duncan, Andrew Paul. "The analysis and application of artificial neural networks for early warning systems in hydrology and the environment." Thesis, University of Exeter, 2014. http://hdl.handle.net/10871/17569.
Full textBouneffouf, Djallel. "DRARS, A Dynamic Risk-Aware Recommender System." Phd thesis, Institut National des Télécommunications, 2013. http://tel.archives-ouvertes.fr/tel-01026136.
Full textFanciulli, Matteo. "Forecast sull'impatto della crescita esponenziale della tecnologia nel mondo del lavoro e nella società." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2016.
Find full textvan, Merriënboer Bart. "Sequence-to-sequence learning for machine translation and automatic differentiation for machine learning software tools." Thèse, 2018. http://hdl.handle.net/1866/21743.
Full textAskari, Hemmat Reyhane. "SLA violation prediction : a machine learning perspective." Thèse, 2016. http://hdl.handle.net/1866/18754.
Full textMokaddem, Mouna. "Learning a graph made of boolean function nodes : a new approach in machine learning." Thèse, 2016. http://hdl.handle.net/1866/18763.
Full textIn this document we present a novel approach in machine learning for classification. The framework we propose is based on boolean circuits, more specifically the classifier produced by our algorithm has that form. Using bits and boolean gates enable the learning algorithm and the classifier to use very efficient boolean vector operations. The accuracy of the classifier we obtain with our framework compares very favourably with those produced by conventional techniques, both in terms of efficiency and accuracy. Furthermore, the framework can be used in a context where information privacy is a necessity, for example we can classify private data. This can be done because computation can be performed only through boolean circuits as encrypted data is quantized in bits. Moreover, assuming that the classifier was trained, it can then be easily implemented on FPGAs (i.e., Field-programmable gate array) as those circuits are also based on logic gates and bitwise operations. Therefore, our model can be easily integrated in real-time classification systems.
Chapados, Nicolas. "Sequential Machine learning Approaches for Portfolio Management." Thèse, 2009. http://hdl.handle.net/1866/3578.
Full textThis thesis considers a number of approaches to make machine learning algorithms better suited to the sequential nature of financial portfolio management tasks. We start by considering the problem of the general composition of learning algorithms that must handle temporal learning tasks, in particular that of creating and efficiently updating the training sets in a sequential simulation framework. We enumerate the desiderata that composition primitives should satisfy, and underscore the difficulty of rigorously and efficiently reaching them. We follow by introducing a set of algorithms that accomplish the desired objectives, presenting a case-study of a real-world complex learning system for financial decision-making that uses those techniques. We then describe a general method to transform a non-Markovian sequential decision problem into a supervised learning problem using a K-best paths search algorithm. We consider an application in financial portfolio management where we train a learning algorithm to directly optimize a Sharpe Ratio (or other risk-averse non-additive) utility function. We illustrate the approach by demonstrating extensive experimental results using a neural network architecture specialized for portfolio management and compare against well-known alternatives. Finally, we introduce a functional representation of time series which allows forecasts to be performed over an unspecified horizon with progressively-revealed information sets. By virtue of using Gaussian processes, a complete covariance matrix between forecasts at several time-steps is available. This information is put to use in an application to actively trade price spreads between commodity futures contracts. The approach delivers impressive out-of-sample risk-adjusted returns after transaction costs on a portfolio of 30 spreads.
Gidel, Gauthier. "Multi-player games in the era of machine learning." Thesis, 2020. http://hdl.handle.net/1866/24800.
Full textAmong all the historical board games played by humans, the game of go was considered one of the most difficult to master by a computer program [Van Den Heriket al., 2002]; Until it was not [Silver et al., 2016]. This odds-breaking break-through [Müller, 2002, Van Den Herik et al., 2002] came from a sophisticated combination of Monte Carlo tree search and machine learning techniques to evaluate positions, shedding light upon the high potential of machine learning to solve games. Adversarial training, a special case of multiobjective optimization, is an increasingly useful tool in machine learning. For example, two-player zero-sum games are important for generative modeling (GANs) [Goodfellow et al., 2014] and mastering games like Go or Poker via self-play [Silver et al., 2017, Brown and Sandholm,2017]. A classic result in Game Theory states that convex-concave games always have an equilibrium [Neumann, 1928]. Surprisingly, machine learning practitioners successfully train a single pair of neural networks whose objective is a nonconvex-nonconcave minimax problem while for such a payoff function, the existence of a Nash equilibrium is not guaranteed in general. This work is an attempt to put learning in games on a firm theoretical foundation. The first contribution explores minimax theorems for a particular class of nonconvex-nonconcave games that encompasses generative adversarial networks. The proposed result is an approximate minimax theorem for two-player zero-sum games played with neural networks, including WGAN, StarCrat II, and Blotto game. Our findings rely on the fact that despite being nonconcave-nonconvex with respect to the neural networks parameters, the payoff of these games are concave-convex with respect to the actual functions (or distributions) parametrized by these neural networks. The second and third contributions study the optimization of minimax problems, and more generally, variational inequalities in the context of machine learning. While the standard gradient descent-ascent method fails to converge to the Nash equilibrium of simple convex-concave games, there exist ways to use gradients to obtain methods that converge. We investigate several techniques such as extrapolation, averaging and negative momentum. We explore these techniques experimentally by proposing a state-of-the-art (at the time of publication) optimizer for GANs called ExtraAdam. We also prove new convergence results for Extrapolation from the past, originally proposed by Popov [1980], as well as for gradient method with negative momentum. The fourth contribution provides an empirical study of the practical landscape of GANs. In the second and third contributions, we diagnose that the gradient method breaks when the game’s vector field is highly rotational. However, such a situation may describe a worst-case that does not occur in practice. We provide new visualization tools in order to exhibit rotations in practical GAN landscapes. In this contribution, we show empirically that the training of GANs exhibits significant rotations around Local Stable Stationary Points (LSSP), and we provide empirical evidence that GAN training converges to a stable stationary point, which is a saddle point for the generator loss, not a minimum, while still achieving excellent performance.
Dauphin, Yann. "Advances in scaling deep learning algorithms." Thèse, 2015. http://hdl.handle.net/1866/13710.
Full textTrabelsi, Amine. "Configuration et exploitation d'une machine émotionnelle." Thèse, 2010. http://hdl.handle.net/1866/4566.
Full textThis work explores the feasibility of equipping computers with the ability to predict, in a context of a human computer interaction, the probable user’s emotion and its intensity for a wide variety of emotion-eliciting situations. More specifically, an online framework, the Emotional Machine, is developed enabling computers to «understand» situations using OCC model of emotion and to predict user’s reaction by combining refined versions of Artificial Neural Network and k Nearest Neighbours algorithms. An empirical procedure including a web-based anonymous questionnaire for data acquisition was designed to provide the chosen machine learning algorithms with a consistent knowledge and to test the application’s recognition performance. Results from the empirical investigation show that the proposed Emotional Machine is capable of producing accurate predictions. Such an achievement may encourage future using of our framework for automated emotion recognition in various application fields.
Mastropietro, Olivier. "Deep Learning for Video Modelling." Thèse, 2017. http://hdl.handle.net/1866/20192.
Full textJean, Sébastien. "From Word Embeddings to Large Vocabulary Neural Machine Translation." Thèse, 2015. http://hdl.handle.net/1866/13421.
Full textIn this thesis, we examine some properties of word embeddings and propose a technique to handle large vocabularies in neural machine translation. We first look at a well-known analogy task and examine the effect of position-dependent weights, the choice of combination function and the impact of supervised learning. We then show that simple embeddings learnt with translational contexts can match or surpass the state of the art on the TOEFL synonym detection task and on the recently introduced SimLex-999 word similarity gold standard. Finally, motivated by impressive results obtained by small-vocabulary (30,000 words) neural machine translation embeddings on some word similarity tasks, we present a GPU-friendly approach to increase the vocabulary size by more than an order of magnitude. Despite originally being developed for obtaining the embeddings only, we show that this technique actually works quite well on actual translation tasks, especially for English to French (WMT'14).
Gupta, Gunshi. "Look-ahead meta-learning for continual learning." Thesis, 2020. http://hdl.handle.net/1866/24315.
Full textThe continual learning problem involves training models with limited capacity to perform well on a set of an unknown number of sequentially arriving tasks. This setup can of- ten see a learning system undergo catastrophic forgetting, when learning a newly seen task causes interference on the learning progress of old tasks. While recent work has shown that meta-learning has the potential to reduce interference between old and new tasks, the current training procedures tend to be either slow or offline, and sensitive to many hyper-parameters. In this work, we propose Look-ahead MAML (La-MAML), a fast optimisation-based meta- learning algorithm for online-continual learning, aided by a small episodic memory. This is achieved by realising the equivalence of a multi-step MAML objective to a time-aware con- tinual learning objective adopted in prior work. The equivalence leads to the formulation of an intuitive algorithm that we call Continual-MAML (C-MAML), employing continual meta- learning to optimise a model to perform well across a series of changing data distributions. By additionally incorporating the modulation of per-parameter learning rates in La-MAML, our approach provides a more flexible and efficient way to mitigate catastrophic forgetting compared to conventional prior-based methods. This modulation also has connections to prior work on meta-descent, which we identify as an important direction of research to de- velop better optimizers for continual learning. In experiments conducted on real-world visual classification benchmarks, La-MAML achieves performance superior to other replay-based, prior-based and meta-learning based approaches for continual learning. We also demonstrate that it is robust, and more scalable than many recent state-of-the-art approaches.
Shabanian, Samira. "Bidirectional Helmholtz Machines." Thèse, 2016. http://hdl.handle.net/1866/16181.
Full textEfficient unsupervised training and inference in deep generative models remains a challenging problem. One basic approach, called Helmholtz machine, involves training a top-down directed generative model together with a bottom-up auxiliary model used for approximate inference. Recent results indicate that better generative models can be obtained with better approximate inference procedures. Instead of improving the inference procedure, we here propose a new model, the bidirectional Helmholtz machine, which guarantees that the top-down and bottom-up distributions can efficiently invert each other. We achieve this by interpreting both the top-down and the bottom-up directed models as approximate inference distributions and by defining the model distribution to be the geometric mean of these two. We present a lower-bound for the likelihood of this model and we show that optimizing this bound regularizes the model so that the Bhattacharyya distance between the bottom-up and top-down approximate distributions is minimized. This approach results in state of the art generative models which prefer significantly deeper architectures while it allows for orders of magnitude more efficient approximate inference. Moreover, we introduce a deep generative model for semi-supervised learning problems based on BiHM models.
Gagnon, Louis-Guillaume. "Searching for supersymmetry using deep learning with the ATLAS detector." Thesis, 2020. http://hdl.handle.net/1866/24811.
Full textThe Standard Model of particle physics (SM) is a fundamental theory of nature whose validity has been extensively confirmed by experiments. However, some theoretical and experimental problems subsist, which motivates searches for alternative theories to supersede it. Supersymmetry (SUSY), which associate new fundamental particles to each SM particle, is one of the best-motivated such theory and could solve some of the biggest outstanding problems with the SM. For example, many SUSY scenarios predict stable neutral particles that could explain observations of dark matter in the universe. The discovery of SUSY would also represent a huge step towards a unified theory of the universe. Searches for SUSY are at the heart of the experimental program of the ATLAS collaboration, which exploits a state-of-the-art particle detector installed at the Large Hadron Collider (LHC) at CERN in Geneva. The probability to observe many supersymmetric particles went up when the LHC ramped up its collision energy to 13~TeV, the highest ever achieved in laboratory, but so far no evidence for SUSY has been recorded by current searches, which are mostly based on well-known simple techniques such as counting experiments. This thesis documents the implementation of a novel deep learning-based approach using only the four-momenta of selected physics objects, and its application to the search for supersymmetric particles using the full ATLAS 2015-2018 dataset. Motivated by naturalness considerations as well as by the problem of dark matter, the search focuses on finding evidence for supersymmetric partners of the gluon (the gluino), third generation quarks (the stop and the sbottom), and gauge bosons (the neutralino). Many recently introduced physics-specific machine learning developments are employed, such as directly using detector-recorded energies and momenta of produced particles instead of first deriving a restricted set of physically motivated variables and parametrizing the classification model with the masses of the particles searched for, which allows optimal sensitivity for all mass hypothesis. This method improves the statistical significance of the search by up to 85 times that of the previous ATLAS analysis for some mass hypotheses, after accounting for the luminosity difference. No significant excesses above the SM background are recorded. Gluino masses below 2.45 TeV and neutralino masses below 1.7 TeV are excluded at the 95% confidence level, greatly increasing the previous limit on two simplified models of gluino pair production with off-shell stops and sbottoms, respectively.
Prato, Gabriele. "Compression in Sequence to Sequence Learning for Natural Language Processing." Thèse, 2019. http://hdl.handle.net/1866/23787.
Full textIn this work, we propose a near lossless method for encoding long sequences of texts as well as all of their sub-sequences into feature rich representations. We test our method on sentiment analysis and show good performance across all sub-sentence and sentence embeddings. This work also demonstrates the use of knowledge distillation and quantization to compress the original Transformer model [Vaswani et al., 2017] for the translation task. We are, to the best of our knowledge, the first to show that 8-bit quantization of the weights of the Transformer can achieve the same BLEU score as the full-precision model. Furthermore, when we combine knowledge distillation with weight quantization, we can train smaller Transformer networks and achieve up to 12.59x compression while losing only 2.51 BLEU off the baseline on the WMT 2014 English-to-French translation task. Chapter 1 introduces machine learning concepts for natural language processing which are essential to understanding both papers presented in this thesis. Chapter 2 and 3 cover each paper respectively, before finally concluding with chapter 4.
Cyr-Cronier, Jessica. "Effets d’âge et de sexe sur la synchronisation de l’EEG en sommeil : analyses multivariées par apprentissage machine." Thèse, 2017. http://hdl.handle.net/1866/20534.
Full textKahya, Emre Onur. "Identifying electrons with deep learning methods." Thesis, 2020. http://hdl.handle.net/1866/25101.
Full textThis thesis is about applying the tools of Machine Learning to an important problem of experimental particle physics: identifying signal electrons after proton-proton collisions at the Large Hadron Collider. In Chapters 1, we provide some information about the Large Hadron Collider and explain why it was built. We give further details about one of the biggest detectors in the Large Hadron Collider, the ATLAS. Then we define what electron identification task is, as well as the importance of solving it. Finally, we give detailed information about our dataset that we use to solve the electron identification task. In Chapters 2, we give a brief introduction to fundamental principles of machine learning. Starting with the definition and types of different learning tasks, we discuss various ways to represent inputs. Then we present what to learn from the inputs as well as how to do it. And finally, we look at the problems that would arise if we “overdo” learning. In Chapters 3, we motivate the choice of the architecture to solve our task, especially for the parts that have sequential images as inputs. We then present the results of our experiments and show that our model performs much better than the existing algorithms that the ATLAS collaboration currently uses. Finally, we discuss future directions to further improve our results. In Chapter 4, we discuss two concepts: out of distribution generalization and flatness of loss surface. We claim that the algorithms, that brings a model into a wide flat minimum of its training loss surface, would generalize better for out of distribution tasks. We give the results of implementing two such algorithms to our dataset and show that it supports our claim. Finally, we end with our conclusions.
Delalleau, Olivier. "Apprentissage machine efficace : théorie et pratique." Thèse, 2012. http://hdl.handle.net/1866/8669.
Full textDespite constant progress in terms of available computational power, memory and amount of data, machine learning algorithms need to be efficient in how they use them. Although minimizing cost is an obvious major concern, another motivation is to attempt to design algorithms that can learn as efficiently as intelligent species. This thesis tackles the problem of efficient learning through various papers dealing with a wide range of machine learning algorithms: this topic is seen both from the point of view of computational efficiency (processing power and memory required by the algorithms) and of statistical efficiency (n umber of samples necessary to solve a given learning task).The first contribution of this thesis is in shedding light on various statistical inefficiencies in existing algorithms. Indeed, we show that decision trees do not generalize well on tasks with some particular properties (chapter 3), and that a similar flaw affects typical graph-based semi-supervised learning algorithms (chapter 5). This flaw is a form of curse of dimensionality that is specific to each of these algorithms. For a subclass of neural networks, called sum-product networks, we prove that using networks with a single hidden layer can be exponentially less efficient than when using deep networks (chapter 4). Our analyses help better understand some inherent flaws found in these algorithms, and steer research towards approaches that may potentially overcome them. We also exhibit computational inefficiencies in popular graph-based semi-supervised learning algorithms (chapter 5) as well as in the learning of mixtures of Gaussians with missing data (chapter 6). In both cases we propose new algorithms that make it possible to scale to much larger datasets. The last two chapters also deal with computational efficiency, but in different ways. Chapter 7 presents a new view on the contrastive divergence algorithm (which has been used for efficient training of restricted Boltzmann machines). It provides additional insight on the reasons why this algorithm has been so successful. Finally, in chapter 8 we describe an application of machine learning to video games, where computational efficiency is tied to software and hardware engineering constraints which, although often ignored in research papers, are ubiquitous in practice.
Trofimov, Assya. "Étude des signatures géniques dans un contexte d’expériences de RNA- Seq." Thèse, 2017. http://hdl.handle.net/1866/20417.
Full textXu, Ge Ya. "Balancing signals for semi-supervised sequence learning." Thèse, 2019. http://hdl.handle.net/1866/23792.
Full textLes réseaux neuronaux récurrents (RNN) sont des modèles puissants qui ont obtenu des réalisations exceptionnelles dans de nombreuses tâches d’apprentissage séquentiel. Malgré leurs réalisations, les modèles RNN sou˙rent encore de longues séquences pendant l’entraî-nement. C’est parce que l’erreur se propage en arrière de la sortie vers les couches d’entrée transportant des signaux de gradient, et avec une longue séquence d’entrée, des problèmes comme la disparition et l’explosion des gradients peuvent survenir. Cette thèse passe en revue de nombreuses études actuelles et architectures existantes conçues pour contour-ner les problèmes de dépendance à long terme de la rétropropagation dans le temps (BPTT). Nous nous concentrons principalement sur la méthode proposée par cite Trinh2018 qui utilise une méthode d’apprentissage semi-supervisée pour atténuer les problèmes de dépendance à long terme dans BPTT. Malgré les bons résultats obtenus avec le modèle de cite Trinh2018, nous suggérons que le modèle peut être encore amélioré avec une manière plus systématique d’équilibrer les signaux auxiliaires. Dans cette thèse, nous présentons notre article - emph RNNs with Private and Shared Representations for Semi-Supervised Learning - qui est actuellement en cours de révision pour AAAI-2019. Nous propo-sons une architecture RNN semi-supervisée avec des représentations privées et partagées explicitement conçues qui régule le flux de gradient de la tâche auxiliaire à la tâche principale.
Bordes, Florian. "Learning to sample from noise with deep generative models." Thèse, 2017. http://hdl.handle.net/1866/19370.
Full textMachine learning and specifically deep learning has made significant breakthroughs in recent years concerning different tasks. One well known application of deep learning is computer vision. Tasks such as detection or classification are nearly considered solved by the community. However, training state-of-the-art models for such tasks requires to have labels associated to the data we want to classify. A more general goal is, similarly to animal brains, to be able to design algorithms that can extract meaningful features from data that aren’t labeled. Unsupervised learning is one of the axes that try to solve this problem. In this thesis, I present a new way to train a neural network as a generative model capable of generating quality samples (a task akin to imagining). I explain how by starting from noise, it is possible to get samples which are close to the training data. This iterative procedure is called Infusion training and is a novel approach to learning the transition operator of a generative Markov chain. In the first chapter, I present some background about machine learning and probabilistic models. The second chapter presents generative models that inspired this work. The third and last chapter presents and investigates our novel approach to learn a generative model with Infusion training.
Breuleux, Olivier. "Échantillonnage dynamique de champs markoviens." Thèse, 2009. http://hdl.handle.net/1866/4316.
Full textOne of the most active topics of research in unsupervised learning is the Boltzmann machine --- particularly the Restricted Boltzmann Machine or RBM. In order to train, evaluate or exploit such models, one has to draw samples from it. Two recent algorithms, Fast Persistent Contrastive Divergence (FPCD) and Herding aim to improve sampling during training. In particular, herding gives up on obtaining a point estimate of the RBM's parameters, rather defining the model's distribution with a dynamical system guided by training samples. We generalize these ideas in order to obtain algorithms capable of exploiting the probability distribution defined by a pre-trained RBM, by sampling from it, without needing to make use of the training set. We present three methods: Sample Penalization, based on a theoretical argument as well as FPCD and Herding using constant statistics for their positive phases. These methods define dynamical systems producing samples with the right statistics and we evaluate them using non-parametric density estimation. We show that these methods mix substantially better than Gibbs sampling, which is the conventional sampling method used for RBMs.
Lemieux, Simon. "Espaces de timbre générés par des réseaux profonds convolutionnels." Thèse, 2011. http://hdl.handle.net/1866/6294.
Full textThis thesis presents a novel way of modelling timbre using machine learning algorithms. More precisely, we have attempted to build a timbre space by extracting audio features using deep-convolutional Boltzmann machines. We first present an overview of machine learning with an emphasis on convolutional Boltzmann machines as well as models from which they are derived. We also present a summary of the literature relevant to timbre spaces and highlight their limitations, such as the small number of timbres used to build them. To address this problem, we have developed a sound generation tool that can generate as many sounds as we wish. At the system's core are plug-ins that are parameterizable and that we can combine to create a virtually infinite range of sounds. We use it to build a massive randomly generated timbre dataset that is made up of real and synthesized instruments. We then train deep-convolutional Boltzmann machines on those timbres in an unsupervised way and use the produced feature space as a timbre space. The timbre space we obtain is a better space than a similar space built using MFCCs. We consider it as better in the sense that the distance between two timbres in that space is more similar to the one perceived by a human listener. However, we are far from reaching the performance of a human. We finish by proposing possible improvements that could be tried to close our performance gap.
Desjardins, Guillaume. "Improving sampling, optimization and feature extraction in Boltzmann machines." Thèse, 2013. http://hdl.handle.net/1866/10550.
Full textDespite the current widescale success of deep learning in training large scale hierarchical models through supervised learning, unsupervised learning promises to play a crucial role towards solving general Artificial Intelligence, where agents are expected to learn with little to no supervision. The work presented in this thesis tackles the problem of unsupervised feature learning and density estimation, using a model family at the heart of the deep learning phenomenon: the Boltzmann Machine (BM). We present contributions in the areas of sampling, partition function estimation, optimization and the more general topic of invariant feature learning. With regards to sampling, we present a novel adaptive parallel tempering method which dynamically adjusts the temperatures under simulation to maintain good mixing in the presence of complex multi-modal distributions. When used in the context of stochastic maximum likelihood (SML) training, the improved ergodicity of our sampler translates to increased robustness to learning rates and faster per epoch convergence. Though our application is limited to BM, our method is general and is applicable to sampling from arbitrary probabilistic models using Markov Chain Monte Carlo (MCMC) techniques. While SML gradients can be estimated via sampling, computing data likelihoods requires an estimate of the partition function. Contrary to previous approaches which consider the model as a black box, we provide an efficient algorithm which instead tracks the change in the log partition function incurred by successive parameter updates. Our algorithm frames this estimation problem as one of filtering performed over a 2D lattice, with one dimension representing time and the other temperature. On the topic of optimization, our thesis presents a novel algorithm for applying the natural gradient to large scale Boltzmann Machines. Up until now, its application had been constrained by the computational and memory requirements of computing the Fisher Information Matrix (FIM), which is square in the number of parameters. The Metric-Free Natural Gradient algorithm (MFNG) avoids computing the FIM altogether by combining a linear solver with an efficient matrix-vector operation. The method shows promise in that the resulting updates yield faster per-epoch convergence, despite being slower in terms of wall clock time. Finally, we explore how invariant features can be learnt through modifications to the BM energy function. We study the problem in the context of the spike & slab Restricted Boltzmann Machine (ssRBM), which we extend to handle both binary and sparse input distributions. By associating each spike with several slab variables, latent variables can be made invariant to a rich, high dimensional subspace resulting in increased invariance in the learnt representation. When using the expected model posterior as input to a classifier, increased invariance translates to improved classification accuracy in the low-label data regime. We conclude by showing a connection between invariance and the more powerful concept of disentangling factors of variation. While invariance can be achieved by pooling over subspaces, disentangling can be achieved by learning multiple complementary views of the same subspace. In particular, we show how this can be achieved using third-order BMs featuring multiplicative interactions between pairs of random variables.
Fréchette, Nicolas. "Segmentation automatique de la fibrose pulmonaire sur images de tomodensitométrie en radio-oncologie." Thèse, 2019. http://hdl.handle.net/1866/23812.
Full textPulmonary fibrosis is an interstitial lung disease characterized by an irreversible production of scarring tissue. Pulmonary fibrosis has a particularly poor prognosis, with a mean survival after diagnosis lower than many cancers. This pathology was recently identified as a risk for complication following radiation therapy treatments. Pulmonary toxicity can lead to severe conditions that compromise the benefits provided by radiation therapy, making pulmonary fibrosis a relative contraindication to treatments. Manual segmentation of fibrosis on computed tomography (CT) images is a difficult task that can involve many experts for a single patient. The aim of this project is to perform automatic segmentation of pulmonary fibrosis on CT images. Fully convolutional neural networks were developed and implemented to automatically assign lung tissues. For an input CT slice, every lung voxel is assigned a tissue in a single inference. Parameters optimization was performed in a supervised and semi-supervised manner by minimizing variants of the cross-entropy between the prediction and manual annotations produced by experts. The dataset employed consists of high resolution CT scans and delineations made by radiologists and radiation oncologists. Predicted segmentation maps were compared with manual segmentations to validate the tissues assigned by the convolutional networks. Results suggest that radiation oncology applications could be developed. Possible applications include pulmonary fibrosis screening prior to treatment planning and assessment of fibrosis progression during and post-treatment.
Fu, Min. "FPGA-based object detection using classification circuits." Thèse, 2015. http://hdl.handle.net/1866/12507.
Full textIn the machine learning area, classification is a process of mapping a new observation to a certain category. Classifiers which implement classification algorithms have been studied widely over the past decades. Traditional classifiers are based on algorithms such as SVM and neural nets, and are usually run by software on CPUs which cause the system to suffer low performance and high power consumption. Although GPUs can be used to accelerate the computation of some classifiers, its high power consumption prevents the technology from being implemented on portable devices such as embedded systems or wearable hardware. To make a lightweight classification system, classifiers should be able to run on a more compact hardware system instead of a group of CPUs/GPUs, and classifiers themselves should be optimized to fit that hardware. In this thesis, we explore the implementation of a novel classifier on a FPGA-based hardware platform. The classifier, devised by Alain Tapp (Université de Montréal), is based on a large amount of look-up tables that form tree-structured circuits to do classification tasks. The FPGA appears to be a tailor-made component to implement this classifier with its rich resources of look-up tables and the highly parallel architecture. Our work shows that a single FPGA can implement multiple classifiers to do classification on high definition images at a very high speed.
Lacaille, Philippe. "Analyzing the benefits of communication channels between deep learning models." Thèse, 2018. http://hdl.handle.net/1866/22128.
Full textPezeshki, Mohammad. "Towards deep semi supervised learning." Thèse, 2016. http://hdl.handle.net/1866/18343.
Full textTouré, Fodé. "Évaluation et amélioration du rendement de la formation en entreprise : vers une démarche basée sur la gestion des processus d’affaires." Thèse, 2014. http://hdl.handle.net/1866/11414.
Full textAlamian, Golnoush. "Investigation of neural activity in Schizophrenia during resting-state MEG : using non-linear dynamics and machine-learning to shed light on information disruption in the brain." Thesis, 2020. http://hdl.handle.net/1866/25254.
Full textPsychiatric disorders affect nearly a quarter of the world’s population. These typically bring about debilitating behavioural, functional and/or cognitive problems, for which the underlying neural mechanisms are poorly understood. These symptoms can significantly reduce the quality of life of affected individuals, impact those close to them, and bring on an economic burden on society. Hence, targeting the baseline neurophysiology associated with psychopathologies, by identifying more robust biomarkers, would improve the development of effective treatments. The first goal of this thesis is thus to contribute to a better characterization of neural dynamic alterations in mental health illnesses, specifically in schizophrenia and mood disorders. Accordingly, the first chapter of this thesis presents two systematic literature reviews, which investigate the resting-state changes in brain connectivity in schizophrenia, depression and bipolar disorder patients. Great strides have been made in neuroimaging research in identifying alterations in functional connectivity. However, these two reviews reveal a gap in the knowledge about the temporal basis of the neural mechanisms involved in the disruption of information integration in these pathologies, particularly in schizophrenia. Therefore, the second goal of this thesis is to characterize the baseline temporal neural alterations of schizophrenia. We present two studies for which we hypothesize that the resting temporal dysconnectivity could serve as a key biomarker in schizophrenia. These studies explore temporal integration deficits in schizophrenia by quantifying neural alterations of scale-free dynamics using resting-state magnetoencephalography (MEG) data. Specifically, we use (1) long-range temporal correlation (LRTC) analysis on oscillatory activity and (2) multifractal analysis on arrhythmic brain activity. In addition, we develop classification models (based on supervised machine-learning) to detect the cortical and sub-cortical features that allow for a robust division of patients and healthy controls. Given that these studies are based on MEG spontaneous brain activity, recorded at rest with either eyes-open or eyes-closed, we then explored the possibility of finding a distinctive feature that would combine both types of resting-state recordings. Thus, the third study investigates whether alterations in spectral amplitude between eyes-open and eyes-closed conditions can be used as a possible marker for schizophrenia. Overall, the three studies show changes in the scale-free dynamics of schizophrenia patients at rest that suggest a deterioration of the temporal processing of information in patients, which might relate to their cognitive and behavioural symptoms. The multimodal approach of this thesis, combining MEG, non-linear analyses and machine-learning, improves the characterization of the resting spatiotemporal neural organization of schizophrenia patients and healthy controls. Our findings provide new evidence for the temporal dysconnectivity hypothesis in schizophrenia. The results extend on previous studies by characterizing scale-free properties of deep brain structures and applying advanced non-linear metrics that are underused in the field of psychiatry. The results of this thesis contribute significantly to the identification of novel biomarkers in schizophrenia and show the importance of clarifying the temporal properties of altered intrinsic neural dynamics. Moreover, the presented studies offer a methodological framework that can be extended to other psychopathologies, such as depression.
Langlois, Dansereau Christian. "Dealing with heterogeneity in the prediction of clinical diagnosis." Thèse, 2017. http://hdl.handle.net/1866/20491.
Full textHamel, Philippe. "Apprentissage de représentations musicales à l'aide d'architectures profondes et multiéchelles." Thèse, 2012. http://hdl.handle.net/1866/8678.
Full textMachine learning (ML) is an important tool in the field of music information retrieval (MIR). Many MIR tasks can be solved by training a classifier over a set of features. For MIR tasks based on music audio, it is possible to extract features from the audio with signal processing techniques. However, some musical aspects are hard to extract with simple heuristics. To obtain richer features, we can use ML to learn a representation from the audio. These learned features can often improve performance for a given MIR task. In order to learn interesting musical representations, it is important to consider the particular aspects of music audio when building learning models. Given the temporal and spectral structure of music audio, deep and multi-scale representations are particularly well suited to represent music. This thesis focuses on learning representations from music audio. Deep and multi-scale models that improve the state-of-the-art for tasks such as instrument recognition, genre recognition and automatic annotation are presented.
Almahairi, Amjad. "Advances in deep learning with limited supervision and computational resources." Thèse, 2018. http://hdl.handle.net/1866/23434.
Full textDeep neural networks are the cornerstone of state-of-the-art systems for a wide range of tasks, including object recognition, language modelling and machine translation. In the last decade, research in the field of deep learning has led to numerous key advances in designing novel architectures and training algorithms for neural networks. However, most success stories in deep learning heavily relied on two main factors: the availability of large amounts of labelled data and massive computational resources. This thesis by articles makes several contributions to advancing deep learning, specifically in problems with limited or no labelled data, or with constrained computational resources. The first article addresses sparsity of labelled data that emerges in the application field of recommender systems. We propose a multi-task learning framework that leverages natural language reviews in improving recommendation. Specifically, we apply neural-network-based methods for learning representations of products from review text, while learning from rating data. We demonstrate that the proposed method can achieve state-of-the-art performance on the Amazon Reviews dataset. The second article tackles computational challenges in training large-scale deep neural networks. We propose a conditional computation network architecture which can adaptively assign its capacity, and hence computations, across different regions of the input. We demonstrate the effectiveness of our model on visual recognition tasks where objects are spatially localized within the input, while maintaining much lower computational overhead than standard network architectures. The third article contributes to the domain of unsupervised learning with the generative adversarial networks paradigm. We introduce a flexible adversarial training framework, in which not only the generator converges to the true data distribution, but also the discriminator recovers the relative density of the data at the optimum. We validate our framework empirically by showing that the discriminator is able to accurately estimate the true energy of data while obtaining state-of-the-art quality of samples. Finally, in the fourth article, we address the problem of unsupervised domain translation. We propose a model which can learn flexible, many-to-many mappings across domains from unpaired data. We validate our approach on several image datasets, and we show that it can be effectively applied in semi-supervised learning settings.
Pelland-Goulet, Pénélope. "Mesure de l'attention visuo-spatiale dans l'espace et le temps par les potentiels reliés aux événements (PRÉ)." Thesis, 2020. http://hdl.handle.net/1866/24397.
Full textEvent related potentials (ERP) are commonly used as a method of measuring visual attention. ERP components such as N2pc and P3 are largely considered as markers of attention deployment. In order to investigate the possibility of predicting the locus and the presence or absence of attention, a spatial cueing task was used. A cue indicated one of the four locations on which subjects had to direct their attention. The spatial cue was exclusively symbolic, implying that attention had to be oriented voluntarily. The analysis of the EEG signal which was measured as subjects carried out the task was performed using machine learning. An SVM (Support Vector Machine) classifier was used to predict the presence or absence of attention at one location, using the EEG signal associated with targets and distractors. A decoding accuracy of 75% (p < 0,001) was achieved for this classification, with a chance level of 50%. A DSVM (Dendrogram SVM) was used to predict the precise locus of attention using the EEG signal linked to targets only. In this classification problem, a decoding accuracy of 51,7% (p < 0,001) was achieved, with a chance level of 25%. These results suggest that it is possible to distinguish the locus of attention from ERPs in a +/- 0,4 degrees of visual angle space with decoding accuracies considerably above chance.
Levade, Inès. "Évolution intra-hôte de Vibrio cholerae et interactions avec le microbiome intestinal." Thesis, 2020. http://hdl.handle.net/1866/25268.
Full textCholera is an acute diarrhoeal disease that remains a global threat to public health in countries where access to safe water and adequate sanitation cannot be guaranteed. Vibrio cholerae, the bacterial pathogen responsible for this disease, can cause a range of symptoms in infected individuals, from intense diarrhea leading to severe dehydration, to asymptomatic carriage of the bacteria. Although our understanding of cholera on a macro-epidemiological scale has been considerably improved by the development of high-throughput sequencing techniques and by advances in bacterial genomics, no studies have yet been conducted to characterize its evolution at the scale of infected individuals. Furthermore, the role of asymptomatic carriers in an epidemic and the reason behind the absence of symptoms in these infected individuals remains unknown. The main objective of this thesis is therefore to characterize the genomic diversity of V. cholerae at the level of individuals and households, but also to evaluate the potential role of the gut microbiome in the susceptibility to contract this acute enteric disease and to present severe symptoms. First, we characterize the genomic diversity of colonies isolated from symptomatic patients. The whole genome sequencing of strains from patients in Bangladesh and Haiti reveals that this diversity is detectable in the form of point mutations within hosts, but remains limited. Much of the variation detected within patients appears to be due to the gain and loss of phages and plasmids within the V. cholerae population, with occasional exchanges between the pathogen and other commensal members of the gut microbiota. These results challenge the commonly accepted assumption that V. cholerae infections are predominantly clonal, and confirm that horizontal gene transfer is an important factor in the evolution of V. cholerae. In addition, our results show that some of these variants may also have a phenotypic effect, for example by impacting biofilm formation, and can be selected within infected individuals. Next, we apply a combination of whole genome sequencing and metagenomic approaches to improve the detection of intra-host variants, both in symptomatic patients and in asymptomatic carriers. Our study shows that the metagenomic approach offers a better resolution in the detection of the diversity in the microbial population, but remains difficult to apply in asymptomatic patients, due to the low number of V. cholerae cells in these individuals. Overall, we find that the level of diversity within the intra-host bacterial population is similar between symptomatic and asymptomatic patients. We also detect the presence of hypermutator strains in some patients. In addition, while mutations in patients with hypermutator phenotypes did not appear to be driven by selection, signs of parallel evolution are detected in patients with fewer mutations, suggesting adaptive mechanisms within the host. Our results underline the power of metagenomics combined with whole genome sequencing to characterize intra-host diversity in acute cholera infection, but also in asymptomatic carriers, while identifying for the first time an hypermutator phenotype in infected patients. Finally, we are interested in factors related to susceptibility to the disease and related to the severity of symptoms. Based on a recent study using 16S rRNA amplicon sequencing to show the potential link between the intestinal microbiome and susceptibility to V. cholerae infection, our study uses metagenomic sequencing methods on the same samples from this previous study to characterize the taxonomic and functional profiles of the gut microbiome of household contacts exposed to V. cholerae. Samples are collected prior to infection of these household contacts, and used to identify predictors of symptomatic disease. Using a machine learning algorithm, we can identify species, gene families and metabolic pathways in the microbiome at the time of exposure to V. cholerae to detect potential biomarkers correlated with risk of infection and symptom severity. Our results show that the use of metagenomic sequencing improves the precision and accuracy of predictions compared to 16S rRNA amplicon sequencing. Our analyses also predict disease severity, although with greater uncertainty than the prediction of infection. Bacterial taxa from the genera Prevotella and Bifidobacterium have been identified as potential markers of protection against infection, as well as genes involved in iron metabolism. Our results highlight the power of metagenomics to predict disease progression and identify specific species and genes that could be involved in experimental tests to study the mechanisms related to the microbiome explaining potential protection against cholera.
Almousli, Hani. "Recognition of Facial Expressions with Autoencoders and Convolutional-Nets." Thèse, 2013. http://hdl.handle.net/1866/10688.
Full textHumans communicate via different types of channels: words, voice, body gesture, emotions …etc. For this reason, implementing these channels in computers is inevitable to make them interact intelligently with humans. Using a webcam and a microphone, computers should figure out what we want to tell from our voice, gesture and face emotions. In this thesis we are interested in figuring human emotions from their images or video in order to use that later in different applications. The thesis starts by giving an introduction to machine learning and some of the models and algorithms we used like multilayer perceptron, convolutional neural networks, autoencoders and finally report the results of applying these models on several facial emotion expression datasets. We moreover concentrate on studying different kinds of autoencoders (Denoising Autoencoder , Contractive Autoencoder, …etc.) and identify some limitations like the possibility of obtaining filters co-adaptation and undesirably smooth spectral curve and we investigate new ideas to address these problems. We also overcome the limitations of training autoencoders in a purely unsupervised manner, i.e. without using any knowledge of task we ultimately want to solve (such as predicting class labels) and develop a new semi-supervised training criterion which exploits the knowledge of the few labeled data to train the autoencoder together with a large amount of unlabeled data in order to learn a representation better suited for the classification task, and obtain better classification performance. Finally, we describe the general pipeline for our emotion detection system and suggest new ideas for future work.
van, Beurden Louis. "Comparaison de systèmes de traduction automatique pour la post édition des alertes météorologique d'Environnement Canada." Thesis, 2019. http://hdl.handle.net/1866/23791.
Full textThe purpose of this paper is to determine the strategy for the automatic translation of weather warnings produced by Environment Canada, which requires the least post-editing effort by the proofreaders of the Translation Bureau. We will begin by developing a bilingual corpus of weather warnings representative of this task. Then, this data will be used to compare the performance of different approaches of machine translation, translation memory configurations and hybrid systems. We will compare the results of these models with the system WATT, the latest system provided by RALI for Environment Canada, as well as with the industry systems GoogleTranslate and DeepL. Finaly, we will study an automatic post-edition system.
Ferreira, Guerra Steve. "Une procédure de sélection automatique de la discrétisation optimale de la ligne du temps pour des méthodes longitudinales d’inférence causale." Thèse, 2017. http://hdl.handle.net/1866/20549.
Full textGotti, Fabrizio. "L'atténuation statistique des surdétections d'un correcteur grammatical symbolique." Thèse, 2012. http://hdl.handle.net/1866/9809.
Full textGrammar checking software sometimes erroneously flags a correct word sequence as an error, a problem we call overdetection in the present study. We describe the devel-opment of a system for identifying and filtering out the overdetections produced by the French grammar checker designed by the firm Druide Informatique. Various fami-lies of classifiers have been trained in a supervised way for 14 types of detections flagged by the grammar checker, using features that capture diverse linguistic phe-nomena (syntactic dependency links, POS tags, word context exploration, etc.), extracted from sentences with and without overdetections. Eight of the 14 classifiers we trained are now part of the latest version of a very popular commercial grammar checker. Moreover, our experiments have shown that statistical language models, SVMs and word sense disambiguation can all contribute to the improvement of these classifiers. This project is a striking illustration of a machine learning component suc-cessfully integrated within a robust, commercial natural language processing application.
Dumoulin, Vincent. "Representation Learning for Visual Data." Thèse, 2018. http://hdl.handle.net/1866/21140.
Full textZhang, Ying. "Sequence to sequence learning and its speech applications." Thèse, 2018. http://hdl.handle.net/1866/21287.
Full textKrueger, David. "Designing Regularizers and Architectures for Recurrent Neural Networks." Thèse, 2016. http://hdl.handle.net/1866/14019.
Full textBibi, Khalil. "Personal information prediction from written texts." Thesis, 2020. http://hdl.handle.net/1866/24308.
Full textAuthorship Attribution (AA) is a field of research that exists since the 60s. It consists of identifying the author of a certain text based on texts with known authors. This is done by extracting features about the writing style and the content of the text. In this master thesis, two sub problems of AA were treated: gender and age classification using a corpus collected from online blogs. In this work, several features were compared using several feature-based algorithms. As well as deep learning methods. For the gender classification task, the best results are the ones obtained by a majority vote system over the outputs of several classifiers. For the age classification task, the best result was obtained using classifier trained over TFIDF.
Dinh, Laurent. "Reparametrization in deep learning." Thèse, 2018. http://hdl.handle.net/1866/21139.
Full textGoodfellow, Ian. "Deep learning of representations and its application to computer vision." Thèse, 2014. http://hdl.handle.net/1866/11674.
Full textKastner, Kyle. "Structured prediction and generative modeling using neural networks." Thèse, 2016. http://hdl.handle.net/1866/18760.
Full textIn this thesis we utilize neural networks to effectively model data with sequential structure. There are many forms of data for which both the order and the structure of the information is incredibly important. The words in this paragraph are one example of this type of data. Other examples include audio, images, and genomes. The work to effectively model this type of ordered data falls within the field of structured prediction. We also present generative models, which attempt to generate data that appears similar to the data which the model was trained on. In Chapter 1, we provide an introduction to data and machine learning. First, we motivate the need for machine learning by describing an expert system built on a customer database. This leads to a discussion of common algorithms, losses, and optimization choices in machine learning. We then progress to describe the basic building blocks of neural networks. Finally, we add complexity to the models, discussing parameter sharing and convolutional and recurrent layers. In the remainder of the document, we discuss several types of neural networks which find common use in both prediction and generative modeling and present examples of their use with audio, handwriting, and images datasets. In Chapter 2, we introduce a variational recurrent neural network (VRNN). Our VRNN is developed with to generate new sequential samples that resemble the dataset that is was trained on. We present models that learned in an unsupervised manner how to generate handwriting, sound effects, and human speech setting benchmarks in performance. Chapter 3 shows a recently developed model called ReNet. In ReNet, intermediate structured outputs from recurrent neural networks are used for object classification. This model shows competitive performance on a number of image recognition tasks, while using an architecture designed to handle structured prediction. In this case, the final model output is only used for simple classification, but follow-up work has expanded to full structured prediction. Lastly, in Chapter 4 we present recent unpublished experiments in sequential audio generation. First we provide background in musical concepts and digital representation which are fundamental to understanding our approach and then introduce a baseline and new research results using our model, RNN-MADE. Next we introduce the concept of raw speech synthesis and discuss our investigation into generation. In our final chapter, we present a brief summary of results and postulate future research directions.
Thibodeau-Laufer, Eric. "Algorithmes d’apprentissage profonds supervisés et non-supervisés: applications et résultats théoriques." Thèse, 2013. http://hdl.handle.net/1866/10689.
Full textThe list of areas affected by machine learning is growing rapidly. As the amount of available training data increases, the development of more powerful learning algorithms is crucial. This thesis consists of three parts: first an overview of the basic concepts of machine learning and the details necessary for training neural networks, models that lend themselves well to deep architectures. The second part presents an application of machine learning to online video games, and a performance measurement method when using these models as decision policies. Finally, the third section presents theoretical results for unsupervised training of deep architectures. Video games are a particularly fertile area for machine learning: it is easy to accumulate large amounts of data, and many tasks are possible. Assembling teams of equal skill is a common machine learning application for online games. The first paper compares different learning algorithms against deep neural networks applied to the prediction of match balance in online games. We then present a simulation based method to evaluate the resulting models used as decision policies for online matchmaking. Following this we present a new method to train generative models. Theoretical results indicate that it is possible to train by backpropagation unsupervised models that can generate samples following the data’s true distribution. This is a relevant result in the context of the recent literature investigating the properties of autoencoders as generative models. These results are supported with preliminary quantitative results and some qualitative experiments.
Wood, Sean. "Non-negative matrix decomposition approaches to frequency domain analysis of music audio signals." Thèse, 2009. http://hdl.handle.net/1866/3769.
Full textWe study the application of unsupervised matrix decomposition algorithms such as Non-negative Matrix Factorization (NMF) to frequency domain representations of music audio signals. These algorithms, driven by a given reconstruction error function, learn a set of basis functions and a set of corresponding coefficients that approximate the input signal. We compare the use of three reconstruction error functions when NMF is applied to monophonic and harmonized musical scales: least squares, Kullback-Leibler divergence, and a recently introduced “phase-aware” divergence measure. Novel supervised methods for interpreting the resulting decompositions are presented and compared to previously used methods that rely on domain knowledge. Finally, the ability of the learned basis functions to generalize across musical parameter values including note amplitude, note duration and instrument type, are analyzed. To do so, we introduce two basis function labeling algorithms that outperform the previous labeling approach in the majority of our tests, instrument type with monophonic audio being the only notable exception.