The occurence of chaos in recurrent neural networks is supposed to depend on the architecture and on the synaptic coupling strength. It is studied here for a randomly diluted architecture. By normalizing the variance of synaptic weights, we produce a bifurcation parameter, dependent on this variance and on the slope of the transfer function but independent of the connectivity, that allows a sustained activity and the occurence of chaos when reaching a critical value. Even for weak connectivity and small size, we find numerical results in accordance with the theoretical ones previously established for fully connected infinite sized networks. Moreover the route towards chaos is numerically checked to be a quasi-periodic one, whatever the type of the first bifurcation is (Hopf bifurcation, pitchfork or flip).

A switching between apparently coherent (oscillatory) and stochastic episodes of activity has been observed in responses from cat and monkey visual cortex. We describe the dynamics of these phenomena in two parallel approaches,a phenomenological and a rather microscopic one. On the one hand we analyze neuronal responses in terms of a hidden state model (HSM). The parameters of this model are extracted directly from experimental spiketrains. They characterize the underlying dynamics as well as the coupling of individual neurons to the network. This phenomenological modelthus provides a new framework for the experimental analysis of network dynamics.

E (0,00), remains in spite of many real (and 459 460 Finnoff imagined)deficiencies the most widely used network training algorithm, and a vast body of literature documents its general applicability and robustness. In this paper we will draw on the highly developed literature of stochastic approximation theory todemonstrate several asymptotic properties of simple backpropagation.

Rep. Germany Abstract Many techniques for model selection in the field of neural networks correspond to well established statistical methods. The method of'stopped training', on the other hand, in which an oversized network is trained until the error on a further validation set of examples deteriorates,then training is stopped, is a true innovation, since model selection doesn't require convergence of the training process. Inthis paper we show that this performance can be significantly enhanced by extending the'nonconvergent model selection method' of stopped training to include dynamic topology modifications (dynamic weight pruning) and modified complexity penalty term methods in which the weighting of the penalty term is adjusted during the training process. 1 INTRODUCTION One of the central topics in the field of neural networks is that of model selection. Both the theoretical and practical side of this have been intensively investigated and a vast array of methods have been suggested to perform this task. A widely used class of techniques starts by choosing an'oversized' network architecture then either removing redundant elements based on some measure of saliency (pruning), adding a further term to the cost function penalizing complexity (penalty terms), and finally, observing the error on a further validation set of examples, then stopping training as soon as this performance begins to deteriorate (stopped training).

This is motivated by experimental evidencethat these phenomena may be subserved by the same mechanisms. An important aspect of this model is that ocular dominance segregationcan occur when input activity is both distributed, and positively correlated between the eyes. This allows investigation of the dependence of the pattern of ocular dominance stripes on the degree of correlation between the eyes: it is found that increasing correlation leads to narrower stripes. Experiments are suggested to test whether such behaviour occursin the natural system.

The bootstrap method offers an computation intensive alternative to estimate the predictive distribution for a neural network even if the analytic derivation is intractable. Theavailable asymptotic results show that it is valid for a large number of linear, nonlinear and even nonparametric regression problems. It has the potential tomodel the distribution of estimators to a higher precision than the usual normal asymptotics. It even may be valid if the normal asymptotics fail. However, the theoretical properties of bootstrap procedures for neural networks - especially nonlinear models - have to be investigated more comprehensively.

Recent research on reinforcement learning has focused on algorithms basedon the principles of Dynamic Programming (DP). One of the most promising areas of application for these algorithms isthe control of dynamical systems, and some impressive results have been achieved. However, there are significant gaps between practice and theory. In particular, there are no con vergence proofsfor problems with continuous state and action spaces, or for systems involving nonlinear function approximators (such as multilayer perceptrons). This paper presents research applying DPbased reinforcement learning theory to Linear Quadratic Regulation (LQR),an important class of control problems involving continuous state and action spaces and requiring a simple type of nonlinear function approximator. We describe an algorithm based on Q-Iearning that is proven to converge to the optimal controller for a large class of LQR problems. We also describe a slightly different algorithm that is only locally convergent to the optimal Q-function, demonstrating one of the possible pitfalls of using a nonlinear function approximator with DPbased learning.

The relationships between learning, development and evolution in Nature is taken seriously, to suggest a model of the developmental process whereby the genotypes manipulated by the Genetic Algorithm (GA)might be expressed to form phenotypic neural networks (NNet) that then go on to learn. ONTOL is a grammar for generating polynomialNNets for time-series prediction. Genomes correspond toan ordered sequence of ONTOL productions and define a grammar that is expressed to generate a NNet. The NNet's weights are then modified by learning, and the individual's prediction error is used to determine GA fitness. A new gene doubling operator appears critical to the formation of new genetic alternatives in the preliminary but encouraging results presented.

These are compared with single unit recordings and behavioural data. The firing of CAl place cells is simulated as the (artificial) rat moves in an environment. Thisis the input for a neuronal network whose output, at each theta (0) cycle, is the next direction of travel for the rat. Cells are characterised by the number of spikes fired and the time of firing with respect to hippocampal 0 rhythm. 'Learning' occurs in'on-off' synapses that are switched on by simultaneous pre-and post-synaptic activity.

The action network is embedded in a sensorymotoric systemarchitecture that contains a separate world model. It is continuously fed with short-term predicted spatiotemporal obstacle trajectories, and receives robot state feedback. The action netallows for external switching between alternative planning tasks.It generates goal-directed motor actions - subject to the robot's kinematic and dynamic constraints - such that collisions withmoving obstacles are avoided. Using supervised learning, we distribute examples of the optimal planner mapping over a structure-level adapted parsimonious higher order network. The training database is generated by a Dynamic Programming algorithm. Extensivesimulations reveal, that the local planner mapping is highly nonlinear, but can be effectively and sparsely represented bythe chosen powerful net model. Excellent generalization occurs for unseen obstacle configurations. We also discuss the limitations offeed-forward neurocontrol for growing planning horizons.