Back-propagation learning (Rumelhart, Hinton and Williams, 1986) is a useful research tool but it has a number of undesiderable features such as having the experimenter decide from outside what should be learned. We describe a number of simulations of neural networks that internally generate their own teaching input. The networks generate the teaching input by trasforming the network input through connection weights that are evolved using a form of genetic algorithm. What results is an innate (evolved) capacity not to behave efficiently in an environment but to learn to behave efficiently. The analysis of what these networks evolve to learn shows some interesting results.

: We present a general method, the ffi-test, which establishes functional dependencies given a sequence of measurements. The approach is based on calculating conditional probabilities from vector component distances. Imposing the requirement of continuity of the underlying function, the obtained values of the conditional probabilities carry information on the embedding dimension and variable dependencies. The power of the method is illustrated on synthetic time-series with different time-lag dependencies and noise levels and on the sunspot data. The virtue of the method for preprocessing data in the context of feed-forward neural networks is demonstrated. Also, its applicability for tracking residual errors in output units is stressed. 1 pihong@thep.lu.se 2 carsten@thep.lu.se Introduction The behaviour of a dynamical system is often modeled by analyzing a time series record of certain system variables. Using artificial neural networks (ANN) to model such systems has recently attr...

Abstract — The problem of determining the proper size of an artificial neural network is recognized to be crucial, especially for its practical implications in such important issues as learning and generalization. One popular approach tackling this problem is commonly known as pruning and consists of training a larger than necessary network and then removing unnecessary weights/nodes. In this paper, a new pruning method is developed, based on the idea of iteratively eliminating units and adjusting the remaining weights in such a way that the network performance does not worsen over the entire training set. The pruning problem is formulated in terms of solving a system of linear equations, and a very efficient conjugate gradient algorithm is used for solving it, in the least-squares sense. The algorithm also provides a simple criterion for choosing the units to be removed, which has proved to work well in practice. The results obtained over various test problems demonstrate the effectiveness of the proposed approach. Index Terms — Feedforward neural networks, generalization, hidden neurons, iterative methods, least-squares methods, network pruning, pattern recognition, structure simplification. I.

. The calculation of second derivatives is required by recent training and analyses techniques of connectionist networks, such as the elimination of superfluous weights, and the estimation of confidence intervals both for weights and network outputs. We here review and develop exact and approximate algorithms for calculating second derivatives. For networks with jwj weights, simply writing the full matrix of second derivatives requires O(jwj 2 ) operations. For networks of radial basis units or sigmoid units, exact calculation of the necessary intermediate terms requires of the order of 2h + 2 backward/forward-propagation passes where h is the number of hidden units in the network. We also review and compare three approximations (ignoring some components of the second derivative, numerical differentiation, and scoring). Our algorithms apply to arbitrary activation functions, networks, and error functions (for instance, with connections that skip layers, or radial basis functions, or ...

Dynamical Recurrent Neural Networks (DRNN) (Aussem 1994) are a class of fully recurrent networks obtained by modeling synapses as autoregressive filters. By virtue of their internal dynamic, these networks approximate the underlying law governing the time series by a system of nonlinear difference equations of internal variables. They therefore provide history-sensitive forecasts without having to be explicitly fed with external memory. The model is trained by a local and recursive error propagation algorithm called temporal-recurrent-backpropagation. The efficiency of the procedure benefits from the exponential decay of the gradient terms backpropagated through the adjoint network. We assess the predictive ability of the DRNN model with meteorological and astronomical time series recorded around the candidate observation sites for the future VLT telescope. The hope is that reliable environmental forecasts provided with the model will allow the modern telescopes to be preset, a few hou...

In the usual setting of Machine Learning, classifiers are typically evaluated by estimating their error rate (or equivalently, the classification accuracy) on the test data. However, this makes sense only if all errors have equal (uniform) costs. When the costs of errors differ between each other, the classifiers should be evaluated by comparing the total costs of the errors.

In this article, we review unsupervised neural network learning procedures which can be applied to the task of preprocessing raw data to extract useful features for subsequent classification. The learning algorithms reviewed here are grouped into three sections: information-preserving methods, density estimation methods, and feature extraction methods. Each of these major sections concludes with a discussion of successful applications of the methods to real-world problems.

It is important to remain uncertain, of observation, model and law. For the Fermi Summer School, Criticisms Requested email : lenny@maths.ox.ac.uk, Contents 1

Emnet for n rv rende licentiatafhandling er design af neurale netv rks ltre. Filtre baseret pa neurale netv rk kan ses som udvidelser af det klassiske line re adaptive l-ter rettet mod modellering af uline re sammenh nge. Hovedv gten l gges pa en neural netv rks implementering af den ikke-rekursive, uline re adaptive model med additiv st j. Formalet er at klarl gge en r kke faser forbundet med design af neural netv rks arkitekturer med henblik pa at udf re forskellige \black-box " modellerings opgaver sa som: System identi kation, invers modellering og pr diktion af tidsserier. De v senligste bidrag omfatter: Formulering af en neural netv rks baseret kanonisk lter repr sentation, der danner baggrund for udvikling af et arkitektur klassi kationssystem. I hovedsagen drejer det sig om en skelnen mellem globale og lokale modeller. Dette leder til at en r kke kendte neurale netv rks arkitekturer kan klassi ceres, og yderligere abnes der mulighed for udvikling af helt nye strukturer. I denne sammenh ng ndes en gennemgang af en r kke velkendte arkitekturer. I s rdeleshed l gges der v gt pa behandlingen af multi-lags perceptron neural netv rket.

The gain of a node in a connectionist network is a multiplicative constant that amplifies or attenuates the net input to the node. The objective of this article is to explore the benefits of adaptive gains in back propagation networks. First we show that gradient descent with respect to gain greatly increases learning speed by amplifying those directions in weight space that are successfully chosen by gradient descent on weights. Adpative gains also allow normalization of weight vectors without loss of computational capacity, and we suggest a simple modification of the learning rule that automatically achieves weight normalization. Finally, we describe a method for creating small hidden layers by making hidden node gains compete according to similarities between nodes, with the goal of improved generalization performance. Simulations show that this competition method is more effective than the special case of gain decay. * In press: IEEE Transactions on Systems, Man and Cybernetics. S...