Antibody-Antigen Interface Characterization with Contextualized Pre-trained Embeddings

Artificial Intelligence (AI) has substantially influenced numerous disciplines in recent years. Biology, chemistry, and bioinformatics are among them, with significant advances in protein structure prediction, paratope prediction, protein-protein interactions (PPIs), and antibody-antigen interactions. Understanding PPIs is critical since they are responsible for practically everything living and have several uses in vaccines, cancer, immunology, and inflammatory illnesses. Machine Learning (ML) offers enormous potential for effectively simulating antibody-antigen interactions and improving in-silico optimization of therapeutic antibodies for desired features, including binding activity, stability, and low immunogenicity. This research looks at the use of AI algorithms to better understand antibody-antigen interactions, and it further expands and explains several difficulties encountered in the field. Furthermore, we contribute by presenting a method that outperforms existing state-of-the-art strategies in paratope prediction from sequence data.

Deep learning of representations and its application to computer vision

L’objectif de cette thèse par articles est de présenter modestement quelques étapes du parcours qui mènera (on espère) à une solution générale du problème de l’intelligence artificielle. Cette thèse contient quatre articles qui présentent chacun une différente nouvelle méthode d’inférence perceptive en utilisant l’apprentissage machine et, plus particulièrement, les réseaux neuronaux profonds. Chacun de ces documents met en évidence l’utilité de sa méthode proposée dans le cadre d’une tâche de vision par ordinateur. Ces méthodes sont applicables dans un contexte plus général, et dans certains cas elles on tété appliquées ailleurs, mais ceci ne sera pas abordé dans le contexte de cette de thèse. Dans le premier article, nous présentons deux nouveaux algorithmes d’inférence variationelle pour le modèle génératif d’images appelé codage parcimonieux “spike- and-slab” (CPSS). Ces méthodes d’inférence plus rapides nous permettent d’utiliser des modèles CPSS de tailles beaucoup plus grandes qu’auparavant. Nous démontrons qu’elles sont meilleures pour extraire des détecteur de caractéristiques quand très peu d’exemples étiquetés sont disponibles pour l’entraînement. Partant d’un modèle CPSS, nous construisons ensuite une architecture profonde, la machine de Boltzmann profonde partiellement dirigée (MBP-PD). Ce modèle a été conçu de manière à simplifier d’entraînement des machines de Boltzmann profondes qui nécessitent normalement une phase de pré-entraînement glouton pour chaque couche. Ce problème est réglé dans une certaine mesure, mais le coût d’inférence dans le nouveau modèle est relativement trop élevé pour permettre de l’utiliser de manière pratique. Dans le deuxième article, nous revenons au problème d’entraînement joint de machines de Boltzmann profondes. Cette fois, au lieu de changer de famille de modèles, nous introduisons un nouveau critère d’entraînement qui donne naissance aux machines de Boltzmann profondes à multiples prédictions (MBP-MP). Les MBP-MP sont entraînables en une seule étape et ont un meilleur taux de succès en classification que les MBP classiques. Elles s’entraînent aussi avec des méthodes variationelles standard au lieu de nécessiter un classificateur discriminant pour obtenir un bon taux de succès en classification. Par contre, un des inconvénients de tels modèles est leur incapacité de générer deséchantillons, mais ceci n’est pas trop grave puisque la performance de classification des machines de Boltzmann profondes n’est plus une priorité étant donné les dernières avancées en apprentissage supervisé. Malgré cela, les MBP-MP demeurent intéressantes parce qu’elles sont capable d’accomplir certaines tâches que des modèles purement supervisés ne peuvent pas faire, telles que celle de classifier des données incomplètes ou encore celle de combler intelligemment l’information manquante dans ces données incomplètes. Le travail présenté dans cette thèse s’est déroulé au milieu d’une période de transformations importantes du domaine de l’apprentissage à réseaux neuronaux profonds qui a été déclenchée par la découverte de l’algorithme de “dropout” par Geoffrey Hinton. Dropout rend possible un entraînement purement supervisé d’architectures de propagation unidirectionnel sans être exposé au danger de sur- entraînement. Le troisième article présenté dans cette thèse introduit une nouvelle fonction d’activation spécialement con ̧cue pour aller avec l’algorithme de Dropout. Cette fonction d’activation, appelée maxout, permet l’utilisation de aggrégation multi-canal dans un contexte d’apprentissage purement supervisé. Nous démontrons comment plusieurs tâches de reconnaissance d’objets sont mieux accomplies par l’utilisation de maxout. Pour terminer, sont présentons un vrai cas d’utilisation dans l’industrie pour la transcription d’adresses de maisons à plusieurs chiffres. En combinant maxout avec une nouvelle sorte de couche de sortie pour des réseaux neuronaux de convolution, nous démontrons qu’il est possible d’atteindre un taux de succès comparable à celui des humains sur un ensemble de données coriace constitué de photos prises par les voitures de Google. Ce système a été déployé avec succès chez Google pour lire environ cent million d’adresses de maisons.

Advances in scaling deep learning algorithms

Les algorithmes d'apprentissage profond forment un nouvel ensemble de méthodes puissantes pour l'apprentissage automatique. L'idée est de combiner des couches de facteurs latents en hierarchies. Cela requiert souvent un coût computationel plus elevé et augmente aussi le nombre de paramètres du modèle. Ainsi, l'utilisation de ces méthodes sur des problèmes à plus grande échelle demande de réduire leur coût et aussi d'améliorer leur régularisation et leur optimization. Cette thèse adresse cette question sur ces trois perspectives. Nous étudions tout d'abord le problème de réduire le coût de certains algorithmes profonds. Nous proposons deux méthodes pour entrainer des machines de Boltzmann restreintes et des auto-encodeurs débruitants sur des distributions sparses à haute dimension. Ceci est important pour l'application de ces algorithmes pour le traitement de langues naturelles. Ces deux méthodes (Dauphin et al., 2011; Dauphin and Bengio, 2013) utilisent l'échantillonage par importance pour échantilloner l'objectif de ces modèles. Nous observons que cela réduit significativement le temps d'entrainement. L'accéleration atteint 2 ordres de magnitude sur plusieurs bancs d'essai. Deuxièmement, nous introduisont un puissant régularisateur pour les méthodes profondes. Les résultats expérimentaux démontrent qu'un bon régularisateur est crucial pour obtenir de bonnes performances avec des gros réseaux (Hinton et al., 2012). Dans Rifai et al. (2011), nous proposons un nouveau régularisateur qui combine l'apprentissage non-supervisé et la propagation de tangente (Simard et al., 1992). Cette méthode exploite des principes géometriques et permit au moment de la publication d'atteindre des résultats à l'état de l'art. Finalement, nous considérons le problème d'optimiser des surfaces non-convexes à haute dimensionalité comme celle des réseaux de neurones. Tradionellement, l'abondance de minimum locaux était considéré comme la principale difficulté dans ces problèmes. Dans Dauphin et al. (2014a) nous argumentons à partir de résultats en statistique physique, de la théorie des matrices aléatoires, de la théorie des réseaux de neurones et à partir de résultats expérimentaux qu'une difficulté plus profonde provient de la prolifération de points-selle. Dans ce papier nous proposons aussi une nouvelle méthode pour l'optimisation non-convexe.

Particle Competition and Cooperation in Networks for Semi-Supervised Learning with Concept Drift

Concept drift is a problem of increasing importance in machine learning and data mining. Data sets under analysis are no longer only static databases, but also data streams in which concepts and data distributions may not be stable over time. However, most learning algorithms produced so far are based on the assumption that data comes from a fixed distribution, so they are not suitable to handle concept drifts. Moreover, some concept drifts applications requires fast response, which means an algorithm must always be (re) trained with the latest available data. But the process of labeling data is usually expensive and/or time consuming when compared to unlabeled data acquisition, thus only a small fraction of the incoming data may be effectively labeled. Semi-supervised learning methods may help in this scenario, as they use both labeled and unlabeled data in the training process. However, most of them are also based on the assumption that the data is static. Therefore, semi-supervised learning with concept drifts is still an open challenge in machine learning. Recently, a particle competition and cooperation approach was used to realize graph-based semi-supervised learning from static data. In this paper, we extend that approach to handle data streams and concept drift. The result is a passive algorithm using a single classifier, which naturally adapts to concept changes, without any explicit drift detection mechanism. Its built-in mechanisms provide a natural way of learning from new data, gradually forgetting older knowledge as older labeled data items became less influent on the classification of newer data items. Some computer simulation are presented, showing the effectiveness of the proposed method.

Collaborative filtering and deep learning based recommendation system for cold start items

Recommender system is a specific type of intelligent systems, which exploits historical user ratings on items and/or auxiliary information to make recommendations on items to the users. It plays a critical role in a wide range of online shopping, e-commercial services and social networking applications. Collaborative filtering (CF) is the most popular approaches used for recommender systems, but it suffers from complete cold start (CCS) problem where no rating record are available and incomplete cold start (ICS) problem where only a small number of rating records are available for some new items or users in the system. In this paper, we propose two recommendation models to solve the CCS and ICS problems for new items, which are based on a framework of tightly coupled CF approach and deep learning neural network. A specific deep neural network SADE is used to extract the content features of the items. The state of the art CF model, timeSVD++, which models and utilizes temporal dynamics of user preferences and item features, is modified to take the content features into prediction of ratings for cold start items. Extensive experiments on a large Netflix rating dataset of movies are performed, which show that our proposed recommendation models largely outperform the baseline models for rating prediction of cold start items. The two proposed recommendation models are also evaluated and compared on ICS items, and a flexible scheme of model retraining and switching is proposed to deal with the transition of items from cold start to non-cold start status. The experiment results on Netflix movie recommendation show the tight coupling of CF approach and deep learning neural network is feasible and very effective for cold start item recommendation. The design is general and can be applied to many other recommender systems for online shopping and social networking applications. The solution of cold start item problem can largely improve user experience and trust of recommender systems, and effectively promote cold start items.

Application of Deep Learning techniques in the search for BSM Higgs bosons in the $\mu\mu$ final state in CMS

The Standard Model (SM) of particle physics predicts the existence of a Higgs field responsible for the generation of particles' mass. However, some aspects of this theory remain unsolved, supposing the presence of new physics Beyond the Standard Model (BSM) with the production of new particles at a higher energy scale compared to the current experimental limits. The search for additional Higgs bosons is, in fact, predicted by theoretical extensions of the SM including the Minimal Supersymmetry Standard Model (MSSM). In the MSSM, the Higgs sector consists of two Higgs doublets, resulting in five physical Higgs particles: two charged bosons $H^{\pm}$, two neutral scalars $h$ and $H$, and one pseudoscalar $A$. The work presented in this thesis is dedicated to the search of neutral non-Standard Model Higgs bosons decaying to two muons in the model independent MSSM scenario. Proton-proton collision data recorded by the CMS experiment at the CERN LHC at a center-of-mass energy of 13 TeV are used, corresponding to an integrated luminosity of $35.9\ \text{fb}^{-1}$. Such search is sensitive to neutral Higgs bosons produced either via gluon fusion process or in association with a $\text{b}\bar{\text{b}}$ quark pair. The extensive usage of Machine and Deep Learning techniques is a fundamental element in the discrimination between signal and background simulated events. A new network structure called parameterised Neural Network (pNN) has been implemented, replacing a whole set of single neural networks trained at a specific mass hypothesis value with a single neural network able to generalise well and interpolate in the entire mass range considered. The results of the pNN signal/background discrimination are used to set a model independent 95\% confidence level expected upper limit on the production cross section times branching ratio, for a generic $\phi$ boson decaying into a muon pair in the 130 to 1000 GeV range.

Geometric deep learning for polygonal mesh denoising

Il seguente lavoro si propone come analisi degli operatori convoluzionali che caratterizzano le graph neural networks. ln particolare, la trattazione si divide in due parti, una teorica e una sperimentale. Nella parte teorica vengono innanzitutto introdotte le nozioni preliminari di mesh e convoluzione su mesh. In seguito vengono riportati i concetti base del geometric deep learning, quali le definizioni degli operatori convoluzionali e di pooling e unpooling. Un'attenzione particolare è stata data all'architettura Graph U-Net. La parte sperimentare riguarda l'applicazione delle reti neurali e l'analisi degli operatori convoluzionali applicati al denoising di superfici perturbate a causa di misurazioni imperfette effettuate da scanner 3D.

A deep learning model via thermodynamics and condensed matter physics

Deep Learning architectures give brilliant results in a large variety of fields, but a comprehensive theoretical description of their inner functioning is still lacking. In this work, we try to understand the behavior of neural networks by modelling in the frameworks of Thermodynamics and Condensed Matter Physics. We approach neural networks as in a real laboratory and we measure the frequency spectrum and the entropy of the weights of the trained model. The stochasticity of the training occupies a central role in the dynamics of the weights and makes it difficult to assimilate neural networks to simple physical systems. However, the analogy with Thermodynamics and the introduction of a well defined temperature leads us to an interesting result: if we eliminate from a CNN the "hottest" filters, the performance of the model remains the same, whereas, if we eliminate the "coldest" ones, the performance gets drastically worst. This result could be exploited in the realization of a training loop which eliminates the filters that do not contribute to loss reduction. In this way, the computational cost of the training will be lightened and more importantly this would be done by following a physical model. In any case, beside important practical applications, our analysis proves that a new and improved modeling of Deep Learning systems can pave the way to new and more efficient algorithms.

Spectral methods in geometric deep learning

Questa tesi propone una panoramica sul funzionamento interno delle architetture alla base del deep learning e in particolare del geometric deep learning. Iniziando a discutere dalla storia degli algoritmi di intelligenza artificiale, vengono introdotti i principali costituenti di questi. In seguito vengono approfonditi alcuni elementi della teoria dei grafi, in particolare il concetto di laplaciano discreto e il suo ruolo nello studio del fenomeno di diffusione sui grafi. Infine vengono presentati alcuni algoritmi utilizzati nell'ambito del geometric deep learning su grafi per la classificazione di nodi. I concetti discussi vengono poi applicati nella realizzazione di un'architettura in grado di classficiare i nodi del dataset Zachary Karate Club.

Geometric Deep Learning per il denoising di mesh 3D

La crescente disponibilità di scanner 3D ha reso più semplice l’acquisizione di modelli 3D dall’ambiente. A causa delle inevitabili imperfezioni ed errori che possono avvenire durante la fase di scansione, i modelli acquisiti possono risultare a volte inutilizzabili ed affetti da rumore. Le tecniche di denoising hanno come obiettivo quello di rimuovere dalla superficie della mesh 3D scannerizzata i disturbi provocati dal rumore, ristabilendo le caratteristiche originali della superficie senza introdurre false informazioni. Per risolvere questo problema, un approccio innovativo è quello di utilizzare il Geometric Deep Learning per addestrare una Rete Neurale in maniera da renderla in grado di eseguire efficacemente il denoising di mesh. L’obiettivo di questa tesi è descrivere il Geometric Deep Learning nell’ambito del problema sotto esame.

Dataset Fusion application on deep learning Object Detection task

Application of dataset fusion techniques to an object detection task, involving the use of deep learning as convolutional neural networks, to manage to create a single RCNN architecture able to inference with good performances on two distinct datasets with different domains.

Graph theory and geometric deep learning

La seguente tesi propone un’introduzione al geometric deep learning. Nella prima parte vengono presentati i concetti principali di teoria dei grafi ed introdotta una dinamica di diffusione su grafo, in analogia con l’equazione del calore. A seguire, iniziando dal linear classifier verranno introdotte le architetture che hanno portato all’ideazione delle graph convolutional networks. In conclusione, si analizzano esempi di alcuni algoritmi utilizzati nel geometric deep learning e si mostra una loro implementazione sul Cora dataset, un insieme di dati con struttura a grafo.

A deep learning model for controlling the galvanizing process in a production line

In the industry of steelmaking, the process of galvanizing is a treatment which is applied to protect the steel from corrosion. The air knife effect (AKE) occurs when nozzles emit a steam of air on the surfaces of a steel strip to remove excess zinc from it. In our work we formalized the problem to control the AKE and we implemented, with the R&D dept.of MarcegagliaSPA, a DL model able to drive the AKE. We call it controller. It takes as input the tuple : a tuple of the physical conditions of the process line (t,h,s) with the target value of the zinc coating (c); and generates the expected tuple of (pres and dist) to drive the mechanical nozzles towards the (c). According to the requirements we designed the structure of the network. We collected and explored the data set of the historical data of the smart factory. Finally, we designed the loss function as sum of three components: the minimization between the coating addressed by the network and the target value we want to reach; and two weighted minimization components for both pressure and distance. In our solution we construct a second module, named coating net, to predict the coating of zinc resulting from the AKE when the conditions are applied to the prod. line. Its structure is made by a linear and a deep nonlinear “residual” component learned by empirical observations. The predictions made by the coating nets are used as ground truth in the loss function of the controller. By tuning the weights of the different components of the loss function, it is possible to train models with slightly different optimization purposes. In the tests we compared the regularization of different strategies with the standard one in condition of optimal estimation for both; the overall accuracy is ± 3 g/m^2 dal target for all of them. Lastly, we analyze how the controller modeled the current solutions with the new logic: the sub-optimal values of pres and dist can be optimize of 50% and 20%.

Learning processes and the neural analysis of conditioning

Classical and operant conditioning principles, such as the behavioral discrepancy-derived assumption that reinforcement always selects antecedent stimulus and response relations, have been studied at the neural level, mainly by observing the strengthening of neuronal responses or synaptic connections. A review of the literature on the neural basis of behavior provided extensive scientific data that indicate a synthesis between the two conditioning processes based mainly on stimulus control in learning tasks. The resulting analysis revealed the following aspects. Dopamine acts as a behavioral discrepancy signal in the midbrain pathway of positive reinforcement, leading toward the nucleus accumbens. Dopamine modulates both types of conditioning in the Aplysia mollusk and in mammals. In vivo and in vitro mollusk preparations show convergence of both types of conditioning in the same motor neuron. Frontal cortical neurons are involved in behavioral discrimination in reversal and extinction procedures, and these neurons preferentially deliver glutamate through conditioned stimulus or discriminative stimulus pathways. Discriminative neural responses can reliably precede operant movements and can also be common to stimuli that share complex symbolic relations. The present article discusses convergent and divergent points between conditioning paradigms at the neural level of analysis to advance our knowledge on reinforcement.

Visualization of learning in multilayer perceptron networks using principal component analysis

This paper is concerned with the use of scientific visualization methods for the analysis of feedforward neural networks (NNs). Inevitably, the kinds of data associated with the design and implementation of neural networks are of very high dimensionality, presenting a major challenge for visualization. A method is described using the well-known statistical technique of principal component analysis (PCA). This is found to be an effective and useful method of visualizing the learning trajectories of many learning algorithms such as back-propagation and can also be used to provide insight into the learning process and the nature of the error surface.