Results 1  10
of
147
Second Order Derivatives for Network Pruning: Optimal Brain Surgeon
 Advances in Neural Information Processing Systems 5
, 1993
"... We investigate the use of information from all second order derivatives of the error function to perform network pruning (i.e., removing unimportant weights from a trained network) in order to improve generalization and increase the speed of further training. ..."
Abstract

Cited by 207 (2 self)
 Add to MetaCart
We investigate the use of information from all second order derivatives of the error function to perform network pruning (i.e., removing unimportant weights from a trained network) in order to improve generalization and increase the speed of further training.
Mean Field Theory for Sigmoid Belief Networks
 Journal of Artificial Intelligence Research
, 1996
"... We develop a mean field theory for sigmoid belief networks based on ideas from statistical mechanics. ..."
Abstract

Cited by 147 (13 self)
 Add to MetaCart
(Show Context)
We develop a mean field theory for sigmoid belief networks based on ideas from statistical mechanics.
Immunology for Physicists
 Reviews of Modern Physics
, 1997
"... The immune system is a complex system of cells and molecules that can provide us with a basic defense against pathogenic organisms. Like the nervous system,... ..."
Abstract

Cited by 78 (0 self)
 Add to MetaCart
(Show Context)
The immune system is a complex system of cells and molecules that can provide us with a basic defense against pathogenic organisms. Like the nervous system,...
Estimating the Scene Illumination Chromaticity by Using a Neural Network
 JOURNAL OF THE OPTICAL SOCIETY OF AMERICA A
, 2002
"... ..."
Genetic Algorithms for Gait Synthesis in a Hexapod Robot
 Recent trends in mobile robots. World Scientific
, 1994
"... This paper describes the staged evolution of a complex motor pattern generator (CPG) for the control of the leg movements of a sixlegged walking robot. The CPG is composed of a network of neurons. In contrast to the main stream work in neural networks, the interconnection weights are altered by ..."
Abstract

Cited by 29 (0 self)
 Add to MetaCart
(Show Context)
This paper describes the staged evolution of a complex motor pattern generator (CPG) for the control of the leg movements of a sixlegged walking robot. The CPG is composed of a network of neurons. In contrast to the main stream work in neural networks, the interconnection weights are altered by a Genetic Algorithm (GA), rather than a learning algorithm. Staged evolution is used to improve the convergence rate of the algorithm, thus obtaining rapid evolution of behavior toward a goal set. First, an oscillator for the individual leg movements is evolved. Then, a network of these oscillators is evolved to coordinate the movements of the different legs. In this way, the designer specifies "islands of fitness" on the way to the final goal, rather than using a single fitness function or determining the explicit solution to the control problem. By introducing a staged set of manageable challenges, the algorithm's performance is improved. These techniques may be applicable to othe...
The dow theory: William peter hamilton’s track record reconsidered
 Journal of Finance
, 1998
"... Abstract: Alfred Cowles ’ (1934) test of the Dow Theory apparently provided strong evidence against the ability of the ability of Wall Street’s most famous chartist to forecast the stock market. In this paper, we review Cowles ’ evidence and find that it supports the contrary conclusion — that the D ..."
Abstract

Cited by 24 (0 self)
 Add to MetaCart
Abstract: Alfred Cowles ’ (1934) test of the Dow Theory apparently provided strong evidence against the ability of the ability of Wall Street’s most famous chartist to forecast the stock market. In this paper, we review Cowles ’ evidence and find that it supports the contrary conclusion — that the Dow Theory, as applied by its major practitioner, William Peter Hamilton over the period 1902 to 1929, yielded positive riskadjusted returns. A reanalysis of the Hamilton editorials suggests that timing strategies based upon the Dow Theory yield high Sharpe ratios and positive alphas. For a current version of this paper, please contact:
Learning by Online Gradient Descent
 Journal of Physics A
, 1995
"... We study online gradientdescent learning in multilayer networks analytically and numerically. The training is based on randomly drawn inputs and their corresponding outputs as defined by a target rule. In the thermodynamic limit we derive deterministic differential equations for the order paramete ..."
Abstract

Cited by 20 (2 self)
 Add to MetaCart
We study online gradientdescent learning in multilayer networks analytically and numerically. The training is based on randomly drawn inputs and their corresponding outputs as defined by a target rule. In the thermodynamic limit we derive deterministic differential equations for the order parameters of the problem which allow an exact calculation of the evolution of the generalization error. First we consider a singlelayer perceptron with sigmoidal activation function learning a target rule defined by a network of the same architecture. For this model the generalization error decays exponentially with the number of training examples if the learning rate is sufficiently small. However, if the learning rate is increased above a critical value, perfect learning is no longer possible. For architectures with hidden layers and fixed hiddentooutput weights, such as the parity and the committee machine, we find additional effects related to the existence of symmetries in these problems...
Evolution of Robustness to Noise and Mutation in Gene Expression Dynamics
, 2007
"... ..."
(Show Context)
The simple dynamics of super Turing theories
, 1994
"... This paper reasons about the need to seek for particular kinds of models of computation that imply stronger computability than the classical models. A possible such model, constituting a chaotic dynamical system, is presented. This system, which we term as the analog shift map, when viewed as a comp ..."
Abstract

Cited by 17 (0 self)
 Add to MetaCart
(Show Context)
This paper reasons about the need to seek for particular kinds of models of computation that imply stronger computability than the classical models. A possible such model, constituting a chaotic dynamical system, is presented. This system, which we term as the analog shift map, when viewed as a computational model has superTuring power and is equivalent to neural networks and the class of analog machines. This map may be appropriate to describe idealized physical phenomena. 1.
Realistic Synaptic Inputs For Model Neural Networks
 Network
, 1991
"... An expression is derived relating the input current for a single neuron in a neural network to the firing rates of excitatory and inhibitory inputs synapsing on the dendritic tree of the neuron. Any dendritic geometry and any pattern of synaptic connections can be treated using the techniques presen ..."
Abstract

Cited by 16 (1 self)
 Add to MetaCart
(Show Context)
An expression is derived relating the input current for a single neuron in a neural network to the firing rates of excitatory and inhibitory inputs synapsing on the dendritic tree of the neuron. Any dendritic geometry and any pattern of synaptic connections can be treated using the techniques presented. The input currents calculated, combined with known firing rate functions, allow the effects of synaptic conductance changes along dendritic cables to be included in a meanfield description of network behavior. The shunting effects of inhibitory synaptic conductances provide a solution to the high firing rate problem in neural network models. Published in Network: Comp. Neural Sys. 2:245258 (1991). 0 Research supported by Department of Energy Contract DEAC0276ER03230 and by National Institute of Mental Health grant MH46742. I. Introduction Neural network models are often based on a meanfield approach [1] that uses known properties of single neurons to predict the behavior of lar...