Modeling High-Dimensional Audio Sequences with Recurrent Neural Networks

Cette thèse étudie des modèles de séquences de haute dimension basés sur des réseaux de neurones récurrents (RNN) et leur application à la musique et à la parole. Bien qu'en principe les RNN puissent représenter les dépendances à long terme et la dynamique temporelle complexe propres aux séquences d'intérêt comme la vidéo, l'audio et la langue naturelle, ceux-ci n'ont pas été utilisés à leur plein potentiel depuis leur introduction par Rumelhart et al. (1986a) en raison de la difficulté de les entraîner efficacement par descente de gradient. Récemment, l'application fructueuse de l'optimisation Hessian-free et d'autres techniques d'entraînement avancées ont entraîné la recrudescence de leur utilisation dans plusieurs systèmes de l'état de l'art. Le travail de cette thèse prend part à ce développement. L'idée centrale consiste à exploiter la flexibilité des RNN pour apprendre une description probabiliste de séquences de symboles, c'est-à-dire une information de haut niveau associée aux signaux observés, qui en retour pourra servir d'à priori pour améliorer la précision de la recherche d'information. Par exemple, en modélisant l'évolution de groupes de notes dans la musique polyphonique, d'accords dans une progression harmonique, de phonèmes dans un énoncé oral ou encore de sources individuelles dans un mélange audio, nous pouvons améliorer significativement les méthodes de transcription polyphonique, de reconnaissance d'accords, de reconnaissance de la parole et de séparation de sources audio respectivement. L'application pratique de nos modèles à ces tâches est détaillée dans les quatre derniers articles présentés dans cette thèse. Dans le premier article, nous remplaçons la couche de sortie d'un RNN par des machines de Boltzmann restreintes conditionnelles pour décrire des distributions de sortie multimodales beaucoup plus riches. Dans le deuxième article, nous évaluons et proposons des méthodes avancées pour entraîner les RNN. Dans les quatre derniers articles, nous examinons différentes façons de combiner nos modèles symboliques à des réseaux profonds et à la factorisation matricielle non-négative, notamment par des produits d'experts, des architectures entrée/sortie et des cadres génératifs généralisant les modèles de Markov cachés. Nous proposons et analysons également des méthodes d'inférence efficaces pour ces modèles, telles la recherche vorace chronologique, la recherche en faisceau à haute dimension, la recherche en faisceau élagué et la descente de gradient. Finalement, nous abordons les questions de l'étiquette biaisée, du maître imposant, du lissage temporel, de la régularisation et du pré-entraînement.

Development and Evaluation of Blind Identification Techniques for Nonlinear Systems

Identification and Control of Non‐linear dynamical systems are challenging problems to the control engineers.The topic is equally relevant in communication,weather prediction ,bio medical systems and even in social systems,where nonlinearity is an integral part of the system behavior.Most of the real world systems are nonlinear in nature and wide applications are there for nonlinear system identification/modeling.The basic approach in analyzing the nonlinear systems is to build a model from known behavior manifest in the form of system output.The problem of modeling boils down to computing a suitably parameterized model,representing the process.The parameters of the model are adjusted to optimize a performanace function,based on error between the given process output and identified process/model output.While the linear system identification is well established with many classical approaches,most of those methods cannot be directly applied for nonlinear system identification.The problem becomes more complex if the system is completely unknown but only the output time series is available.Blind recognition problem is the direct consequence of such a situation.The thesis concentrates on such problems.Capability of Artificial Neural Networks to approximate many nonlinear input-output maps makes it predominantly suitable for building a function for the identification of nonlinear systems,where only the time series is available.The literature is rich with a variety of algorithms to train the Neural Network model.A comprehensive study of the computation of the model parameters,using the different algorithms and the comparison among them to choose the best technique is still a demanding requirement from practical system designers,which is not available in a concise form in the literature.The thesis is thus an attempt to develop and evaluate some of the well known algorithms and propose some new techniques,in the context of Blind recognition of nonlinear systems.It also attempts to establish the relative merits and demerits of the different approaches.comprehensiveness is achieved in utilizing the benefits of well known evaluation techniques from statistics. The study concludes by providing the results of implementation of the currently available and modified versions and newly introduced techniques for nonlinear blind system modeling followed by a comparison of their performance.It is expected that,such comprehensive study and the comparison process can be of great relevance in many fields including chemical,electrical,biological,financial and weather data analysis.Further the results reported would be of immense help for practical system designers and analysts in selecting the most appropriate method based on the goodness of the model for the particular context.

A Reinforcement Learning Approach to Economic Dispatch using Neural Networks

This paper presents a Reinforcement Learning (RL) approach to economic dispatch (ED) using Radial Basis Function neural network. We formulate the ED as an N stage decision making problem. We propose a novel architecture to store Qvalues and present a learning algorithm to learn the weights of the neural network. Even though many stochastic search techniques like simulated annealing, genetic algorithm and evolutionary programming have been applied to ED, they require searching for the optimal solution for each load demand. Also they find limitation in handling stochastic cost functions. In our approach once we learn the Q-values, we can find the dispatch for any load demand. We have recently proposed a RL approach to ED. In that approach, we could find only the optimum dispatch for a set of specified discrete values of power demand. The performance of the proposed algorithm is validated by taking IEEE 6 bus system, considering transmission losses

Forecasting Global Temperature Variations by Neural Networks

Global temperature variations between 1861 and 1984 are forecast usingsregularization networks, multilayer perceptrons and linearsautoregression. The regularization network, optimized by stochasticsgradient descent associated with colored noise, gives the bestsforecasts. For all the models, prediction errors noticeably increasesafter 1965. These results are consistent with the hypothesis that thesclimate dynamics is characterized by low-dimensional chaos and thatsthe it may have changed at some point after 1965, which is alsosconsistent with the recent idea of climate change.s

Design and Development of 3-DOF Modular Micro Parallel Kinematic Manipulator

This paper presents the research and development of a 3-legged micro Parallel Kinematic Manipulator (PKM) for positioning in micro-machining and assembly operations. The structural characteristics associated with parallel manipulators are evaluated and the PKMs with translational and rotational movements are identified. Based on these identifications, a hybrid 3-UPU (Universal Joint-Prismatic Joint-Universal Joint) parallel manipulator is designed and fabricated. The principles of the operation and modeling of this micro PKM is largely similar to a normal size Stewart Platform (SP). A modular design methodology is introduced for the construction of this micro PKM. Calibration results of this hybrid 3-UPU PKM are discussed in this paper.

Efficient learning of reactive robot behaviors with a Neural-Q_learning approach

The purpose of this paper is to propose a Neural-Q_learning approach designed for online learning of simple and reactive robot behaviors. In this approach, the Q_function is generalized by a multi-layer neural network allowing the use of continuous states and actions. The algorithm uses a database of the most recent learning samples to accelerate and guarantee the convergence. Each Neural-Q_learning function represents an independent, reactive and adaptive behavior which maps sensorial states to robot control actions. A group of these behaviors constitutes a reactive control scheme designed to fulfill simple missions. The paper centers on the description of the Neural-Q_learning based behaviors showing their performance with an underwater robot in a target following task. Real experiments demonstrate the convergence and stability of the learning system, pointing out its suitability for online robot learning. Advantages and limitations are discussed

Semi-online neural-Q_leaming for real-time robot learning

Reinforcement learning (RL) is a very suitable technique for robot learning, as it can learn in unknown environments and in real-time computation. The main difficulties in adapting classic RL algorithms to robotic systems are the generalization problem and the correct observation of the Markovian state. This paper attempts to solve the generalization problem by proposing the semi-online neural-Q_learning algorithm (SONQL). The algorithm uses the classic Q_learning technique with two modifications. First, a neural network (NN) approximates the Q_function allowing the use of continuous states and actions. Second, a database of the most representative learning samples accelerates and stabilizes the convergence. The term semi-online is referred to the fact that the algorithm uses the current but also past learning samples. However, the algorithm is able to learn in real-time while the robot is interacting with the environment. The paper shows simulated results with the "mountain-car" benchmark and, also, real results with an underwater robot in a target following behavior

Interacting software-agents to support a kind of expert supervision systems

Expert supervision systems are software applications specially designed to automate process monitoring. The goal is to reduce the dependency on human operators to assure the correct operation of a process including faulty situations. Construction of this kind of application involves an important task of design and development in order to represent and to manipulate process data and behaviour at different degrees of abstraction for interfacing with data acquisition systems connected to the process. This is an open problem that becomes more complex with the number of variables, parameters and relations to account for the complexity of the process. Multiple specialised modules tuned to solve simpler tasks that operate under a co-ordination provide a solution. A modular architecture based on concepts of software agents, taking advantage of the integration of diverse knowledge-based techniques, is proposed for this purpose. The components (software agents, communication mechanisms and perception/action mechanisms) are based on ICa (Intelligent Control architecture), software middleware supporting the build-up of applications with software agent features

Data-Driven Text Generation using Neural Networks & Provenance is Complicated and Boring — Is there a solution?

Title: Data-Driven Text Generation using Neural Networks Speaker: Pavlos Vougiouklis, University of Southampton Abstract: Recent work on neural networks shows their great potential at tackling a wide variety of Natural Language Processing (NLP) tasks. This talk will focus on the Natural Language Generation (NLG) problem and, more specifically, on the extend to which neural network language models could be employed for context-sensitive and data-driven text generation. In addition, a neural network architecture for response generation in social media along with the training methods that enable it to capture contextual information and effectively participate in public conversations will be discussed. Speaker Bio: Pavlos Vougiouklis obtained his 5-year Diploma in Electrical and Computer Engineering from the Aristotle University of Thessaloniki in 2013. He was awarded an MSc degree in Software Engineering from the University of Southampton in 2014. In 2015, he joined the Web and Internet Science (WAIS) research group of the University of Southampton and he is currently working towards the acquisition of his PhD degree in the field of Neural Network Approaches for Natural Language Processing. Title: Provenance is Complicated and Boring — Is there a solution? Speaker: Darren Richardson, University of Southampton Abstract: Paper trails, auditing, and accountability — arguably not the sexiest terms in computer science. But then you discover that you've possibly been eating horse-meat, and the importance of provenance becomes almost palpable. Having accepted that we should be creating provenance-enabled systems, the challenge of then communicating that provenance to casual users is not trivial: users should not have to have a detailed working knowledge of your system, and they certainly shouldn't be expected to understand the data model. So how, then, do you give users an insight into the provenance, without having to build a bespoke system for each and every different provenance installation? Speaker Bio: Darren is a final year Computer Science PhD student. He completed his undergraduate degree in Electronic Engineering at Southampton in 2012.

Integration of knowledge-based, qualitative and numeric tools for real time dynamic systems supervision

The proposal presented in this thesis is to provide designers of knowledge based supervisory systems of dynamic systems with a framework to facilitate their tasks avoiding interface problems among tools, data flow and management. The approach is thought to be useful to both control and process engineers in assisting their tasks. The use of AI technologies to diagnose and perform control loops and, of course, assist process supervisory tasks such as fault detection and diagnose, are in the scope of this work. Special effort has been put in integration of tools for assisting expert supervisory systems design. With this aim the experience of Computer Aided Control Systems Design (CACSD) frameworks have been analysed and used to design a Computer Aided Supervisory Systems (CASSD) framework. In this sense, some basic facilities are required to be available in this proposed framework: ·

Real-Time software control of neural stimulation in a closed loop Animat system

A recent area for investigation into the development of adaptable robot control is the use of living neuronal networks to control a mobile robot. The so-called Animat paradigm comprises a neuronal network (the ‘brain’) connected to an external embodiment (in this case a mobile robot), facilitating potentially robust, adaptable robot control and increased understanding of neural processes. Sensory input from the robot is provided to the neuronal network via stimulation on a number of electrodes embedded in a specialist Petri dish (Multi Electrode Array (MEA)); accurate control of this stimulation is vital. We present software tools allowing precise, near real-time control of electrical stimulation on MEAs, with fast switching between electrodes and the application of custom stimulus waveforms. These Linux-based tools are compatible with the widely used MEABench data acquisition system. Benefits include rapid stimulus modulation in response to neuronal activity (closed loop) and batch processing of stimulation protocols.

Detection of Parkinson’s disease tremor-onset with a frequency analysing artificial neural network

Deep Brain Stimulator devices are becoming widely used for therapeutic benefits in movement disorders such as Parkinson's disease. Prolonging the battery life span of such devices could dramatically reduce the risks and accumulative costs associated with surgical replacement. This paper demonstrates how an artificial neural network can be trained using pre-processing frequency analysis of deep brain electrode recordings to detect the onset of tremor in Parkinsonian patients. Implementing this solution into an 'intelligent' neurostimulator device will remove the need for continuous stimulation currently used, and open up the possibility of demand-driven stimulation. Such a methodology could potentially decrease the power consumption of a deep brain pulse generator.

Combining Human and Machine Brains - Practical systems in information & control

In this paper a look is taken at how the use of implant technology can be used to either increase the range of the abilities of a human and/or diminish the effects of a neural illness, such as Parkinson's Disease. The key element is the need for a clear interface linking the human brain directly with a computer. The area of interest here is the use of implant technology, particularly where a connection is made between technology and the human brain and/or nervous system. Pilot tests and experimentation are invariably carried out apriori to investigate the eventual possibilities before human subjects are themselves involved. Some of the more pertinent animal studies are discussed here. The paper goes on to describe human experimentation, in particular that carried out by the author himself, which led to him receiving a neural implant which linked his nervous system bi-directionally with the internet. With this in place neural signals were transmitted to various technological devices to directly control them. In particular, feedback to the brain was obtained from the fingertips of a robot hand and ultrasonic (extra) sensory input. A view is taken as to the prospects for the future, both in the near term as a therapeutic device and in the long term as a form of enhancement.

Closely coupled behaviour in networked robot systems

Multiple cooperating robot systems may be required to take up a closely coupled configuration in order to perform a task. An example is extended baseline stereo (EBS), requiring that two robots must establish and maintain for a certain period of time a constrained kinematic relationship to each other. In this paper we report on the development of a networked robotics framework for modular, distributed robot systems that supports the creation of such configurations. The framework incorporates a query mechanism to locate modules distributed across the two robot systems. The work presented in this paper introduces special mechanisms to model the kinematic constraint and its instantiation. The EBS configuration is used as a case study and experimental implementation to demonstrate the approach.