The Infona portal uses cookies, i.e. strings of text saved by a browser on the user's device. The portal can access those files and use them to remember the user's data, such as their chosen settings (screen view, interface language, etc.), or their login data. By using the Infona portal the user accepts automatic saving and using this information for portal operation purposes. More information on the subject can be found in the Privacy Policy and Terms of Service. By closing this window the user confirms that they have read the information on cookie usage, and they accept the privacy policy and the way cookies are used by the portal. You can change the cookie settings in your browser.
We discuss properties of stable recurrent artificial neural systems. In particular, we focus on the “detailed balance” net which is a stable recurrent asymmetric net for Hebbian learning, and on its applications. Furthermore, we summarize results obtained for a hybrid approach containing Hebbian as well as back propagation terms.
A long range Ising Neural Network is considered, where p patterns have been stored according to either the Hebb rule or a modification of it which stores patterns with different weights. By performing computer simulations, the size of the basins of attraction of pure and spurious memories is evaluated in the absence of noise, as a function of the load parameter of the system. It is found that the...
First, we study the effects of introducing training noise on the retrieval behaviours of dilute attractor neural networks. We found that, in general, training noise enhances associativity, but also reduces the attractor overlap. At a narrow range of storage levels, however, the system exhibits re-entrant retrieval behaviour on increasing training noise. Secondly, we consider optimization of...
Many learning paradigms for neural networks are based on the optimization of some measure of performance. The learning behavior of these networks then depends on the topography of the corresponding “fitness landscape” and on the chosen optimization method. In this paper, we apply different learning strategies to a number of Boolean problems and analyze how the topography of the fitness landscape is...
These lectures review the basic aspects of learning in simple neural networks and outlines how one can study learning as a statistical dynamical process, stressing the role of phase transitions.
An extension of the Hopfield model is studied, whereby a certain number of ‘privileged’ patterns are ‘marked’, during the learning process. The replica trick is used in studying the retrieval properties of such networks. The relevance of the extension to other biological processes is also discussed.
The replica method of E. Gardner is used to calculate several properties of the perceptron with maximal stability. In particular results on learning times for biased patterns, the problem of generalization, unsupervised learning and a fast learning algorithm (Adatron) are reported.
One of the main research topics within neuron-like networks is related to learning techniques. Competitive learning has got an special interest among them, because a great network automation is achieved with it, ie, autonomously and without explicit indication of the correct output patterns, the network extracts general features that can be used in order to cluster a set of patterns. In this paper,...
This paper reviews recent work on neural network models with sign constrained weights. The review will cover learning rules for neural networks with sign constrained weights, the optimal storage capacity of these models and the effect on the dynamics of retrieval, non-retrieval and uniform attractors.
A model of electrotonic neurons is presented which allows computer simulation of a physiologically realistic neural network, such as that found in nematodes. The undulatory locomotion of Caenorhabditis elegans is investigated by solving the equations of motion of a segmented model of the discretized body taking into account all internal and external forces. The spatio-temporal muscle excitation patterns...
Set the date range to filter the displayed results. You can set a starting date, ending date or both. You can enter the dates manually or choose them from the calendar.