Introduction 1.1 Learning in Recurrent Networks Connectionist networks having feedback connections are interesting for a number of reasons. Biological neural networks are highly recurrently connected, and many authors have studied recurrent network models of various types of perceptual and memory processes. The general property making such networks interesting and potentially useful is that they manifest highly nonlinear dynamical behavior. One such type of dynamical behavior that has received much attention is that of settling to a fixed stable state, but probably of greater importance both biologically and from an engineering viewpoint are time-varying behaviors. Here we consider algorithms for training recurrent networks to perform temporal supervised learning tasks, in which the specification of desired behavior is in the form of specific examples of input and desired output trajectories. One example of such a task is sequence classification, where

We have seen three important trends develop in the last five years in speech recognition. First, heterogeneous parameter sets that mix absolute spectral information with dynamic, or time-derivative, spectral information, have become common. Second, similariry transform techniques, often used to normalize and decor-relate parameters in some computationally inexpensive way, have become popular. Third, the signal parameter estimation problem has merged with the speech recognition process so that more sophisticated statistical models of the signal’s spectrum can be estimated in a closed-loop manner. In this paper, we review the signal processing components of these algorithms. These al-gorithms are presented as part of a unified view of the signal parameterization problem in which there are three major tasks: measurement, transformation, and statistical modeling. This paper is by no means a comprehensive survey of all possible techniques of signal modeling in speech recognition. There are far too many algorithms in use today to make an exhaustive survey feasible (and cohesive). Instead, this paper is meant to serve as a tutorial on signal processing in state-of-the-art speech recognition systems and to review those techniques most commonly used. In keeping with this goal, a complete mathematical description of each algorithm has been included in the paper.

HINE Parameters X0, X1, ....Xp Output E0, E1,....Ep Error Desired Output D0, D1,...Dp Y0, Y1,...Yp Input w w0 w1 AT&T Laboratories (c) COST FUNCTION Output E0, E1,....Ep Error Desired Output D0, D1,...Dp Y0, Y1,...Yp X0, X1, ....Xp Input Parameters w B R A COMPUTING THE GRADIENT WITH BACKPROPAGATION O = A(I1, I2) dI1 = dO ¶ A ¶ I1 dI2 = dO ¶ A ¶ I2 - The learning machine is composed of modules (e.g. layers) - Each module can do two things: 1- compute its outputs from its inputs (FPROP) 2- compute gradient vectors at its inputs from gradient vectors at its outputs (BPROP) A O, dO I1, dI1 I2, dI2 AT&T Laboratories (c) AN INTERESTING SPECIAL CASE: MULTILAYER NETWORKS X0, X1, ....Xp Output Desired Output D0, D1,...Dp Y0, Y1,...Yp Input || D - Y || 2 2 1 WX F() WX F() Mean Square Error Parameters (weights + biases) w Weight matrix E0, E1,....Ep Sigmoids + Biase

We consider problems of sequence processing and propose a solution based on a discrete state model in order to represent past context. Weintroduce a recurrent connectionist architecture having a modular structure that associates a subnetwork to each state. The model has a statistical interpretation we call Input/Output Hidden Markov Model (IOHMM). It can be trained by the EM or GEM algorithms, considering state trajectories as missing data, which decouples temporal credit assignment and actual parameter estimation. The model presents similarities to hidden Markov models (HMMs), but allows us to map input se-quences to output sequences, using the same processing style as recurrent neural networks. IOHMMs are trained using a more discriminant learning paradigm than HMMs, while potentially taking advantage of the EM algorithm. We demonstrate that IOHMMs are well suited for solving grammatical inference problems on a benchmark problem. Experimental results are presented for the seven Tomita grammars, showing that these adaptive models can attain excellent generalization.

Supervised Learning in Multi-Layered Neural Networks (MLNs) has been recently proposed through the well-known Backpropagation algorithm. This is a gradient method which can get stuck in local minima, as simple examples can show. In this paper, some conditions on the network architecture and the learning environment are proposed which ensure the convergence of the Backpropagation algorithm. It is proven in particular that the convergence holds if the classes are linearly-separable. In this case, the experience gained in several experiments shows that MLNs exceed perceptrons in generalization to new examples. Index Terms- Multi-Layered Networks, learning environment, Backpropagation, pattern recognition, linearly-separable classes. I. Introduction Supervised learning in Multi-Layered Networks can be accomplished thanks to Backpropagation (BP ) ([19, 25, 31]). Its application to several different subjects [25], and, particularly, to pattern recognition ([3, 6, 8, 20, 27, 29]), has bee...

INTRODUCTION The ability of multilayer back-propagation networks to learn complex, high-dimensional, nonlinear mappings from large collections of examples makes them obvious candidates for image recognition or speech recognition tasks (see PATTERN RECOGNITION AND NEURAL NETWORKS). In the traditional model of pattern recognition, a hand-designed feature extractor gathers relevant information from the input and eliminates irrelevant variabilities. A trainable classifier then categorizes the resulting feature vectors (or strings of symbols) into classes. In this scheme, standard, fully-connected multilayer networks can be used as classifiers. A potentially more interesting scheme is to eliminate the feature extractor, feeding the network with "raw" inputs (e.g. normalized images), and to rely on backpropagation to turn the first few layers into an appropriate feature extractor. While this can be done with an ordinary fully connected feed-forward network with some success for tasks

A neural network architecture, which models synapses as Finite Impulse Response (FIR) linear filters, is discussed for use in time series prediction. Analysis and methodology are detailed in the context of the Santa Fe Institute Time Series Prediction Competition. Results of the competition show that the FIR network performed remarkably well on a chaotic laser intensity time series. 1 Introduction The goal of time series prediction or forecasting can be stated succinctly as follows: given a sequence y(1); y(2); : : : y(N) up to time N , find the continuation y(N + 1); y(N + 2)::: The series may arise from the sampling of a continuous time system, and be either stochastic or deterministic in origin. The standard prediction approach involves constructing an underlying model which gives rise to the observed sequence. In the oldest and most studied method, which dates back to Yule [1], a linear autoregression (AR) is fit to the data: y(k) = T X n=1 a(n)y(k \Gamma n) + e(k) = y(k) + ...

We present a writer independent system for on-line handwriting recognition which can handle a variety of writing styles including cursive script and hand-print. The input to our system contains the pen trajectory information, encoded as a time-ordered sequence of feature vectors. A Time Delay Neural Network is used to estimate a posteriori probabilities for characters in a word. A Hidden Markov Model segments the word in a way which optimizes the global word score, taking a dictionary into account. A geometrical normalization scheme and a fast but efficient dictionary search are also presented. Trained on 20k words from 59 writers, using a 25k word dictionary we reached a 89% character and 80% word recognition rate on test data from a disjoint set of writers. Keywords: Handwriting Recognition, Neural Networks, Cursive Script, Hidden Markov Models, Dictionary Search. 1 Introduction Pen interfaces should replace advantageously both mouse and keyboard in a variety of situations. Users w...

Statistical techniques based on hidden Markov Models (HMMs) with Gaussian emission densities have dominated signal processing and pattern recognition literature for the past 20 years. However, HMMs trained using maximum likelihood techniques suffer from an inability to learn discriminative information and are prone to overfitting and over-parameterization. Recent work in machine learning has focused on models, such as the support vector machine (SVM), that automatically control generalization and parameterization as part of the overall optimization process. In this paper, we show that SVMs provide a significant improvement in performance on a static pattern classification task based on the Deterding vowel data. We also describe an application of SVMs to large vocabulary speech recognition, and demonstrate an improvement in error rate on a continuous alphadigit task (OGI Aphadigits) and a large vocabulary conversational speech task (Switchboard). Issues related to the development and optimization of an SVM/HMM hybrid system are discussed.

This report is concerned with integrating connectionist networks into a hidden Markov model (HMM) speech recognition system, This is achieved through a statistical understanding of connectionist networks as probability estimators, first elucidated by Herve Bourlard. We review the basis of HMM speech recognition, and point out the possible benefits of incorporating connectionist networks. We discuss some issues necessary to the construction of a connectionist HMM recognition system, and describe the performance of such a system, including evaluations on the DARPA database, in collaboration with Mike Cohen and Horacio Franco of SRI International. In conclusion, we show that a connectionist component improves a state of the art HMM system. ii Part I INTRODUCTION Over the past few years, connectionist models have been widely proposed as a potentially powerful approach to speech recognition (e.g. Makino et al. (1983), Huang et al. (1988) and Waibel et al. (1989)). However, whilst connec...