A major goal of research on networks of neuron-like processing units is to discover efficient learning procedures that allow these networks to construct complex internal representations of their environment. The learning procedures must be capable of modifying the connection strengths in such a way that internal units which are not part of the input or output come to represent important features of the task domain. Several interesting gradient-descent procedures have recently been discovered. Each connection computes the derivative, with respect to the connection strength, of a global measure of the error in the performance of the network. The strength is then adjusted in the direction that decreases the error. These relatively simple, gradient-descent learning procedures work well for small tasks and the new challenge is to find ways of improving their convergence rate and their generalization abilities so that they can be applied to larger, more realistic tasks.

The influence of prior experience on some forms of behavior and cognition is drastically affected by damage to the hippocampal system. However, if the hippocampal system is left intact both during the experience and for a period of time thereafter, subsequent damage can have much less or even no effect. Such findings suggest that memory traces change over time in a way that makes them less dependent on the hippocampal system. This process of change has often been called consolidation. Consolidation is a very gradual process; in humans, it appears to span up to 15 years. This article asks what consolidation is and why it occurs. We take as our point of departure the view that the initial memory trace that results from a relevant experience consists of changes to the strengths of the connections among neurons in the hippocampal system. Bidirectional connections between the neocortex and the hippocampus allow these initial traces to mediate the reinstatement of representations of events o...

It is possible to combine multiple probabilistic models of the same data by multiplying the probabilities together and then renormalizing. This is a very efficient way to model high-dimensional data which simultaneously satisfies many different low-dimensional constraints. Each individual expert model can focus on giving high probability to data vectors that satisfy just one of the constraints. Data vectors that satisfy this one constraint but violate other constraints will be ruled out by their low probability under the other expert models. Training a product of models appears difficult because, in addition to maximizing the probabilities that the individual models assign to the observed data, it is necessary to make the models disagree on unobserved regions of the data space: It is fine for one model to assign a high probability to an unobserved region as long as some other model assigns it a very low probability. Fortunately, if the individual models are tractable there is a fairly efficient way to train a product of models. This training algorithm suggests a biologically plausible way of learning neural population codes.

The error backpropagation learning algorithm (BP) is generally considered biologically implausible because it does not use locally available, activation-based variables. A version of BP that can be computed locally using bi-directional activation recirculation (Hinton & McClelland, 1988) instead of backpropagated error derivatives is more biologically plausible. This paper presents a generalized version of the recirculation algorithm (GeneRec), which overcomes several limitations of the earlier algorithm by using a generic recurrent network with sigmoidal units that can learn arbitrary input/output mappings. However, the contrastiveHebbian learning algorithm (CHL, a.k.a. DBM or mean field learning) also uses local variables to perform error-driven learning in a sigmoidal recurrent network. CHL was derived in a stochastic framework (the Boltzmann machine), but has been extended to the deterministic case in various ways, all of which rely on problematic approximationsand assumptions, le...

This paper describes and motivates six principles for computational cognitive neuroscience models: biological realism, distributed representations, inhibitory competition, bidirectional activation propagation, errordriven task learning, and Hebbian model learning. Although these principles are supported by a number of cognitive, computational, and biological motivations, the prototypical neural network model (a feedforward backpropagation network) incorporates only two of them, and no widely used model incorporates all of them. This paper argues that these principles should be integrated into a coherent overall framework, and discusses some potential synergies and conflicts in doing so.

Computational models in cognitive neuroscience should ideally use biological properties and powerful computational principles to produce behavior consistent with psychological findings. Error-driven backpropagation is computationally powerful, and has proven useful for modeling a range of psychological data, but is not biologically plausible. Several approaches to implementing backpropagation in a biologically plausible fashion converge on the idea of using bidirectional activation propagation in interactive networks to convey error signals. This paper demonstrates two main points about these error-driven interactive networks: (a) they generalize poorly due to attractor dynamics that interfere with the network's ability to systematically produce novel combinatorial representations in response to novel inputs; and (b) this generalization problem can be remedied by adding two widely used mechanistic principles, inhibitory competition and Hebbian learning, that can be independent...

ed on visual and speech data. The ability of the network to automatically generate wavelet codes from natural images is demonstrated. These bear a close resemblance to 2-D Gabor functions, which have previously been used to describe physiological receptive fields, and as a means of producing compact image representations. Keywords: neural networks, unsupervised learning, self-organisation, feature extraction, information theory, redundancy reduction, sparse coding, imaging models, occlusion, image coding, speech coding. Declaration This dissertation is the result of my own original work, except where reference is made to the work of others. No part of it has been submitted for any other university degree or diploma. Its length, including captions, footnotes, appendix and bibliography, is approximately 58000 words. Acknowledgements I would like first and foremost to thank Richard Prager, my supervisor, fo

This Carnegie Symposium celebrates a growing convergence of behavioral and neural approaches to the mechanisms of cognitive change and reflects an overall convergence of behavioral and neural approaches to all aspects of cognition and cognitive development. This article is a part of a corresponding convergence in my own research. In my own formative years as a psychologist, facts about underlying neural mechanisms were considered to be nearly irrelevant to understanding cognition and its development. Where I went to graduate school we studied cognition and cognitive development on the one hand and physiological psychology on the other, and they seemed almost completely disconnected subjects. My own early experimental work stayed strictly on the cognitive side of this huge divide. Things began to change for me in the late 1970’s, with the emergence of connectionist models. They seemed to many of us who were involved in their development to represent a clear step toward bridging the gap between mind and brain (Hinton & Anderson, 1981). We saw information processing as arising from the interactions of vast numbers of neurons. We began to think of propagation of activation among neurons via their synaptic connections as producing cognitive outcomes ranging from perception to comprehension to problem solving, and we thought of changes in the

Although anticipation is an important part of creating believable behaviour, it has had but a secondary role in the field of life-like characters. In this paper, we show how a simple anticipatory mechanism can be used to control the behaviour of a synthetic character implemented as a software agent, without disrupting the user’s suspension of disbelief. We describe the emotivector, an anticipatory mechanism coupled with a sensor, that: (1) uses the history of the sensor to anticipate the next sensor state; (2) interprets the mismatch between the prediction and the sensed value, by computing its attention grabbing potential and associating a basic qualitative sensation with the signal; (3) sends its interpretation along with the signal. When a signal from the sensor reaches the processing module of the agent, it carries recommendations such as: “you should seriously take this signal into consideration, as it is much better than we had expected ” or “just forget about this one, it is as bad as we predicted”. We delineate several strategies to manage several emotivectors at once and show how one of these strategies (meta-anticipation) transparently introduces the concept of uncertainty. Finally, we describe an experiment in which an emotivector-controlled synthetic character interacts with the user in the context of a wordpuzzle game and present the evaluation supporting the adequacy of our approach.

BackprP&&mL666 and contrm866 e Hebbianlearnm6 ar two methods oftr--Pfi5W networ8 with hidden neur&;-- BackprW--PmL6WW computes aner--W signalfor the output neurtm andspr&--j it over the hidden neurW--5 Contrm--85 e Hebbian learnm; involves clamping the outputneurm8 at desirP values, and letting the e#ect sprtm thrtmW feedback connections over the entir networ-- To investigate ther elationship between these twofor& of lear;W;m weconsider a special case in which they ar identical, a multilayer perWfi6;mL withlinear output units, to which weak feedback connections have been added. In this case, the change in networ state caused by clamping the outputneurm6 tur5 out to be the same as the erW-- signal spr&W by backprPfi;mLW656 exceptfor a scalar pr------6&mL This suggests that the functionality of backprP&&mLWWP can ber--&88&m altermLWW ely by a Hebbian-type learP5j algorP5mL which is suitablefor implementation in biological networW5 1