Today, there is a demand for reliable ways to perform automatic modulation recognition of Low Probability of Intercept (LPI) radar signals, not least in the defense industry. This study explores the possibility of performing automatic modulation recognition on these signals through clustering and more specifically how to learn representations of input signals for this task. A semi-supervised approach using a bootstrapped convolutional neural network classifier for representation learning is proposed. A comparison is made between training the representation learner on raw time-series and on spectral representations of the input signals. It is concluded that, overall, the system trained on spectral representations performs better, though both approaches show promise and should be explored further. The proposed system is tested both on known modulation types and on previously unseen modulation types in the task of novelty detection. The results show that the system can successfully identify known modulation types with adjusted mutual information of 0.86 for signal-to-noise ratios ranging from -10 dB to 10 dB. When introducing previously unseen modulations, up to six modulations can be identified with adjusted mutual information above 0.85. Furthermore, it is shown that the system can learn to separate LPI radar signals from telecom signals which are present in most signal environments. / Idag finns ett behov av pålitlig automatiserad modulationsigenkänning (AMR) av Low Probability of Inercept (LPI)-radarsignaler, inte minst hos försvarsindustrin. Denna studie utforskar möjligheten att utföra AMR av dessa signaler genom klustring och mer specifikt hur man bör lära in representationer av signalerna i detta syfte. En halvövervakad inlärningsmetod som använder en klassificerare baserad på faltningsnätverk föreslås. En jämförelse görs mellan ett system som tränar för representationsinlärning på råa tidsserier och ett system som tränar på spektrala representationer av signalerna. Resultaten visar att systemet tränat på spektrala representationer på det stora hela presterar bättre, men båda metoderna visar lovande resultat och bör utforskas vidare. Systemet testas på signaler från både kända och för systemet tidigare okända modulationer i syfte att pröva förmågan att upptäcka nya typer av modulationer. Systemet identifierar kända modulationer med adjusted mutual information på 0.86 i brusnivåer från -10 dB till 10 dB. När tidigare okända modulationer introduceras till systemet ligger adjusted mutual information över 0.85 för upp till sex modulationer. Studien visar dessutom att systemet kan lära sig skilja LPI-radarsignaler från telekommunikationssignaler som är vanliga i de flesta signalmiljöer.

In recent years, machine learning models have gained traction in the field of empirical asset pricing for their risk premium prediction performance. In this thesis, we build upon the work of [1] by first evaluating models similar to their best performing model in a similar fashion, by using the same dataset and measures, and then expanding upon that. We explore the impact of different feature extraction techniques, ranging from simply removing added complex- ity to representation learning techniques such as incremental PCA and autoen- coders. Furthermore, we also introduce recurrent connections with LSTM and combine them with the earlier mentioned representation learning techniques. We significantly outperform [1] in terms of monthly out-of-sample R2, reach- ing a score of over 3%, by using a condensed version of the dataset, without interaction terms and dummy variables, with a feedforward neural network. However, across the board, all of our models fall short in terms of Sharpe ratio. Even though we find that LSTM works better than the benchmark, it does not outperform the feedforward network using the condensed dataset. We reason that this is because the features already contain a lot of temporal information, such as recent price trends. Overall, the autoencoder based models perform poorly. While the linear incremental PCA based models perform better than the nonlinear autoencoder based ones, they still perform worse than the bench- mark. / Under de senaste åren har maskininlärningsmodeller vunnit kredibilitet inom området empirisk tillgångsvärdering för deras förmåga att förutsäga riskpre- mier. I den här uppsatsen bygger vi på [1]s arbetet genom att först implemente- ra modeller som liknar deras bäst presterande modell och utvärdera dem på ett liknande sätt, genom att använda samma data och mått, och sedan bygga vida- re på det. Vi utforskar effekterna av olika variabelextraktionstekniker, allt från att helt enkelt ta bort extra komplexitet till representationsinlärningstekniker som inkrementell PCA och autoencoders. Vidare introducerar vi även LSTM och kombinerar dem med de tidigare nämnda representationsinlärningstekni- kerna. Min bästa modell presterar betydligt bättre än [1]s i termer av månatlig R2 för testdatan, och når ett resultat på över 3%, genom att använda en kompri- merad version av datan, utan interaktionstermer och dummyvariabler, med ett feedforward neuralt nätverk. Men överlag så brister alla mina modeller i ter- mer av Sharpe ratio. Även om LSTM fungerar bättre än riktvärdet, överträffar det inte feedforward-nätverket med den komprimerade datamängden. Vi re- sonerar att detta är på grund av inputvariablerna som redan innehåller en hel del information över tid, som de senaste pristrenderna. Sammantaget presterar de autoencoderbaserade modellerna dåligt. Även om de linjära inkrementell PCA-baserade modellerna presterar bättre än de olinjära autoencoderbaserade modellerna, presterar de fortfarande sämre än riktvärdet.

Recent progress on deep neural networks has shown remarkable action recognition performance from videos. The remarkable performance is often achieved by transfer learning: training a model on a large-scale labeled dataset (source) and then fine-tuning the model on the small-scale labeled datasets (targets). However, existing action recognition models do not always generalize well on new tasks or datasets because of the following two reasons. i) Current action recognition datasets have a spurious correlation between action types and background scene types. The models trained on these datasets are biased towards the scene instead of focusing on the actual action. This scene bias leads to poor generalization performance. ii) Directly testing the model trained on the source data on the target data leads to poor performance as the source, and target distributions are different. Fine-tuning the model on the target data can mitigate this issue. However, manual labeling small- scale target videos is labor-intensive. In this dissertation, I propose solutions to these two problems. For the first problem, I propose to learn scene-invariant action representations to mitigate the scene bias in action recognition models. Specifically, I augment the standard cross-entropy loss for action classification with 1) an adversarial loss for the scene types and 2) a human mask confusion loss for videos where the human actors are invisible. These two losses encourage learning representations unsuitable for predicting 1) the correct scene types and 2) the correct action types when there is no evidence. I validate the efficacy of the proposed method by transfer learning experiments. I trans- fer the pre-trained model to three different tasks, including action classification, temporal action localization, and spatio-temporal action detection. The results show consistent improvement over the baselines for every task and dataset. I formulate human action recognition as an unsupervised domain adaptation (UDA) problem to handle the second problem. In the UDA setting, we have many labeled videos as source data and unlabeled videos as target data. We can use already exist- ing labeled video datasets as source data in this setting. The task is to align the source and target feature distributions so that the learned model can generalize well on the target data. I propose 1) aligning the more important temporal part of each video and 2) encouraging the model to focus on action, not the background scene, to learn domain-invariant action representations. The proposed method is simple and intuitive while achieving state-of-the-art performance without training on a lot of labeled target videos. I relax the unsupervised target data setting to a sparsely labeled target data setting. Then I explore the semi-supervised video action recognition, where we have a lot of labeled videos as source data and sparsely labeled videos as target data. The semi-supervised setting is practical as sometimes we can afford a little bit of cost for labeling target data. I propose multiple video data augmentation methods to inject photometric, geometric, temporal, and scene invariances to the action recognition model in this setting. The resulting method shows favorable performance on the public benchmarks. / Doctor of Philosophy / Recent progress on deep learning has shown remarkable action recognition performance. The remarkable performance is often achieved by transferring the knowledge learned from existing large-scale data to the small-scale data specific to applications. However, existing action recog- nition models do not always work well on new tasks and datasets because of the following two problems. i) Current action recognition datasets have a spurious correlation between action types and background scene types. The models trained on these datasets are biased towards the scene instead of focusing on the actual action. This scene bias leads to poor performance on the new datasets and tasks. ii) Directly testing the model trained on the source data on the target data leads to poor performance as the source, and target distributions are different. Fine-tuning the model on the target data can mitigate this issue. However, manual labeling small-scale target videos is labor-intensive. In this dissertation, I propose solutions to these two problems. To tackle the first problem, I propose to learn scene-invariant action representations to mitigate background scene- biased human action recognition models for the first problem. Specifically, the proposed method learns representations that cannot predict the scene types and the correct actions when there is no evidence. I validate the proposed method's effectiveness by transferring the pre-trained model to multiple action understanding tasks. The results show consistent improvement over the baselines for every task and dataset. To handle the second problem, I formulate human action recognition as an unsupervised learning problem on the target data. In this setting, we have many labeled videos as source data and unlabeled videos as target data. We can use already existing labeled video datasets as source data in this setting. The task is to align the source and target feature distributions so that the learned model can generalize well on the target data. I propose 1) aligning the more important temporal part of each video and 2) encouraging the model to focus on action, not the background scene. The proposed method is simple and intuitive while achieving state-of-the-art performance without training on a lot of labeled target videos. I relax the unsupervised target data setting to a sparsely labeled target data setting. Here, we have many labeled videos as source data and sparsely labeled videos as target data. The setting is practical as sometimes we can afford a little bit of cost for labeling target data. I propose multiple video data augmentation methods to inject color, spatial, temporal, and scene invariances to the action recognition model in this setting. The resulting method shows favorable performance on the public benchmarks.

Extraire une représentation de tous les facteurs de haut niveau de l'état d'un agent à partir d'informations sensorielles de bas niveau est une tâche importante, mais difficile, dans l'apprentissage automatique. Dans ce memoire, nous explorerons plusieurs approches non supervisées pour apprendre ces représentations. Nous appliquons et analysons des méthodes d'apprentissage de représentations non supervisées existantes dans des environnements d'apprentissage par renforcement, et nous apportons notre propre suite d'évaluations et notre propre méthode novatrice d'apprentissage de représentations d'état. Dans le premier chapitre de ce travail, nous passerons en revue et motiverons l'apprentissage non supervisé de représentations pour l'apprentissage automatique en général et pour l'apprentissage par renforcement. Nous introduirons ensuite un sous-domaine relativement nouveau de l'apprentissage de représentations : l'apprentissage auto-supervisé. Nous aborderons ensuite deux approches fondamentales de l'apprentissage de représentations, les méthodes génératives et les méthodes discriminatives. Plus précisément, nous nous concentrerons sur une collection de méthodes discriminantes d'apprentissage de représentations, appelées méthodes contrastives d'apprentissage de représentations non supervisées (CURL). Nous terminerons le premier chapitre en détaillant diverses approches pour évaluer l'utilité des représentations. Dans le deuxième chapitre, nous présenterons un article de workshop dans lequel nous évaluons un ensemble de méthodes d'auto-supervision standards pour les problèmes d'apprentissage par renforcement. Nous découvrons que la performance de ces représentations dépend fortement de la dynamique et de la structure de l'environnement. À ce titre, nous déterminons qu'une étude plus systématique des environnements et des méthodes est nécessaire. Notre troisième chapitre couvre notre deuxième article, Unsupervised State Representation Learning in Atari, où nous essayons d'effectuer une étude plus approfondie des méthodes d'apprentissage de représentations en apprentissage par renforcement, comme expliqué dans le deuxième chapitre. Pour faciliter une évaluation plus approfondie des représentations en apprentissage par renforcement, nous introduisons une suite de 22 jeux Atari entièrement labellisés. De plus, nous choisissons de comparer les méthodes d'apprentissage de représentations de façon plus systématique, en nous concentrant sur une comparaison entre méthodes génératives et méthodes contrastives, plutôt que les méthodes générales du deuxième chapitre choisies de façon moins systématique. Enfin, nous introduisons une nouvelle méthode contrastive, ST-DIM, qui excelle sur ces 22 jeux Atari. / Extracting a representation of all the high-level factors of an agent’s state from level-level sensory information is an important, but challenging task in machine learning. In this thesis, we will explore several unsupervised approaches for learning these state representations. We apply and analyze existing unsupervised representation learning methods in reinforcement learning environments, as well as contribute our own evaluation benchmark and our own novel state representation learning method. In the first chapter, we will overview and motivate unsupervised representation learning for machine learning in general and for reinforcement learning. We will then introduce a relatively new subfield of representation learning: self-supervised learning. We will then cover two core representation learning approaches, generative methods and discriminative methods. Specifically, we will focus on a collection of discriminative representation learning methods called contrastive unsupervised representation learning (CURL) methods. We will close the first chapter by detailing various approaches for evaluating the usefulness of representations. In the second chapter, we will present a workshop paper, where we evaluate a handful of off-the-shelf self-supervised methods in reinforcement learning problems. We discover that the performance of these representations depends heavily on the dynamics and visual structure of the environment. As such, we determine that a more systematic study of environments and methods is required. Our third chapter covers our second article, Unsupervised State Representation Learning in Atari, where we try to execute a more thorough study of representation learning methods in RL as motivated by the second chapter. To facilitate a more thorough evaluation of representations in RL we introduce a benchmark of 22 fully labelled Atari games. In addition, we choose the representation learning methods for comparison in a more systematic way by focusing on comparing generative methods with contrastive methods, instead of the less systematically chosen off-the-shelf methods from the second chapter. Finally, we introduce a new contrastive method, ST-DIM, which excels at the 22 Atari games.

Graphs are often considered an excellent way of modeling complex real-world problems since they allow to capture relationships between items. Because of their ubiquity, graph embedding techniques have occupied research groups, seeking how vertices can be encoded into a low-dimensional latent space, useful to then perform machine learning. Recently Graph Neural Networks (GNN) have dominated the space of embeddings generation due to their inherent ability to encode latent node dependencies. Moreover, the newly introduced Inductive Graph Neural Networks gained much popularity for inductively learning and representing node embeddings through neighborhood aggregate measures. Even when an entirely new node, unseen during training, appears in the graph, it can still be properly represented by its neighboring nodes. Although this approach appears suitable for dynamic graphs, available systems and training methodologies are agnostic of dynamicity and solely rely on re-processing full graph snapshots in batches, an approach that has been criticized for its high computational costs. This work provides a thorough solution to this particular problem via an efficient prioritybased method for selecting rehearsed samples that guarantees low complexity and high accuracy. Finally, a data-parallel inference method has been evaluated at scale using Apache Flink, a data stream processor for real-time predictions on high volume graph data streams. / Molti problemi nel mondo reale possono essere rappresentati come grafi poichè queste strutture dati consentono di modellare relazioni tra elementi. A causa del loro vasto uso, molti gruppi di ricerca hanno tentato di rappresentare i vertici in uno spazio a bassa dimensione, utile per poi poter utilizzare tecniche di apprendimento automatico. Le reti neurali per grafi sono state ampiamente utilizzate per via della loro capacità di codificare dipendenze tra vertici. Le reti neurali induttive recentemente introdotte, inoltre, hanno guadagnato popolarità poichè consentono di generare rappresentazioni di vertici aggregando altri vertici. In questo modo anche un nodo completamente nuovo può comunque essere rappresentato utilizzando i suoi nodi vicini. Sebbene questo approccio sia adatto per grafici dinamici, i sistemi ad oggi disponibili e gli algoritmi di addestramento si basano esclusivamente sulla continua elaborazione di grafi statici, un approccio che è stato criticato per i suoi elevati costi di calcolo. Questa tesi fornisce una soluzione a questo problema tramite un metodo efficiente per l’allenamento di reti neurali induttive basato su un’euristica per la selezione dei vertici. Viene inoltre descritto un metodo per eseguire predizioni in modo scalabile in tempo reale utilizzando Apache Flink, un sistema per l’elaborazione di grandi quantità di flussi di dati in tempo reale. / Grafer anses ofta vara ett utmärkt sätt att modellera komplexa problem i verkligheten eftersom de gör det möjligt att fånga relationer mellan objekt. På grund av deras allestädes närhet har grafinbäddningstekniker sysselsatt forskningsgrupper som undersöker hur hörn kan kodas in i ett lågdimensionellt latent utrymme, vilket är användbart för att sedan utföra maskininlärning. Nyligen har Graph Neural Networks (GNN) dominerat utrymmet för inbäddningsproduktion tack vare deras inneboende förmåga att koda latenta nodberoenden. Dessutom fick de nyinförda induktiva grafiska nervnäten stor popularitet för induktivt lärande och representerande nodbäddningar genom sammanlagda åtgärder i grannskapet. Även när en helt ny nod, osynlig under träning, visas i diagrammet, kan den fortfarande representeras ordentligt av dess angränsande noder. Även om detta tillvägagångssätt tycks vara lämpligt för dynamiska grafer, är tillgängliga system och träningsmetodologier agnostiska för dynamik och förlitar sig bara på att behandla fullständiga ögonblicksbilder i partier, en metod som har kritiserats för dess höga beräkningskostnader. Detta arbete ger en grundlig lösning på detta specifika problem via en effektiv prioriteringsbaserad metod för att välja repeterade prover som garanterar låg komplexitet och hög noggrannhet. Slutligen har en dataparallell inferensmetod utvärderats i skala med Apache Flink, en dataströmprocessor för realtidsprognoser för grafiska dataströmmar med hög volym.

No description available.

No description available.

Nous présentons dans cette thèse un nouveau paradigme pour caractériser la maladie d’Alzheimer à travers l’écriture manuscrite acquise sur tablette graphique. L’état de l’art est dominé par des méthodes qui supposent un comportement unique ou homogène au sein de chaque profil cognitif. Ces travaux exploitent des paramètres cinématiques globaux, sur lesquels ils appliquent des tests statistiques ou des algorithmes de classification pour discriminer les différents profils cognitifs (les patients Alzheimer, les troubles cognitifs légers (« Mild Cognitive impairment » : MCI) et les sujets Contrôle (HC)). Notre travail aborde ces deux limites de la littérature de la façon suivante : premièrement au lieu de considérer un comportement homogène au sein de chaque profil cognitif ou classe (HC, MCI, ES-AD : « Early-Stage Alzheimer Disease »), nous nous sommes affranchis de cette hypothèse (ou contrainte) forte de la littérature. Nous considérons qu’il peut y avoir plusieurs comportements au sein de chaque profil cognitif. Ainsi, nous proposons un apprentissage semi-supervisé pour trouver des groupes homogènes de sujets et analysons l’information contenue dans ces clusters ou groupes sur les profils cognitifs. Deuxièmement, au lieu d’exploiter les paramètres cinématiques globaux (ex : vitesse moyenne, pression moyenne, etc.), nous avons défini deux paramétrisations ou codages : une paramétrisation semi-globale, puis locale en modélisant la dynamique complète de chaque paramètre. L’un de nos résultats importants met en évidence deux clusters majeurs qui sont découverts, l’un dominé par les sujets HC et MCI et l’autre par les MCI et ES-AD, révélant ainsi que les patients atteints de MCI ont une motricité fine qui est proche soit des sujets HC, soit des patients ES-AD. Notre travail montre également que la vitesse prise localement regroupe un ensemble riche des caractéristiques telles que la taille, l’inclinaison, la fluidité et la régularité, et révèle comment ces paramètres spatiotemporels peuvent conjointement caractériser les profils cognitifs. / We present, in this thesis, a novel paradigm for assessing Alzheimer’s disease by analyzing impairment of handwriting (HW) on tablets, a challenging problem that is still in its infancy. The state of the art is dominated by methods that assume a unique behavioral trend for each cognitive profile, and that extract global kinematic parameters, assessed by standard statistical tests or classification models, for discriminating the neuropathological disorders (Alzheimer’s (AD), Mild Cognitive Impairment (MCI)) from Healthy Controls (HC). Our work tackles these two major limitations as follows. First, instead of considering a unique behavioral pattern for each cognitive profile, we relax this heavy constraint by allowing the emergence of multimodal behavioral patterns. We achieve this by performing semi-supervised learning to uncover homogeneous clusters of subjects, and then we analyze how much information these clusters carry on the cognitive profiles. Second, instead of relying on global kinematic parameters, mostly consisting of their average, we refine the encoding either by a semi-global parameterization, or by modeling the full dynamics of each parameter, harnessing thereby the rich temporal information inherently characterizing online HW. Thanks to our modeling, we obtain new findings that are the first of their kind on this research field. A striking finding is revealed: two major clusters are unveiled, one dominated by HC and MCI subjects, and one by MCI and ES-AD, thus revealing that MCI patients have fine motor skills leaning towards either HC’s or ES-AD’s. This thesis introduces also a new finding from HW trajectories that uncovers a rich set of features simultaneously like the full velocity profile, size and slant, fluidity, and shakiness, and reveals, in a naturally explainable way, how these HW features conjointly characterize, with fine and subtle details, the cognitive profiles.

No description available.

No description available.