New approximation method for smooth error backpropagation in a quantron network

Montigny, Simon de

doi:10.1016/j.neunet.2014.07.015

Cited by 4 publications

(2 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The quantron, a structural element designed for artificial neural networks (ANN), has been proposed and analyzed in previous works [1][2][3][4][5]. It has proven capable of solving effectively the benchmark XOR problem [1].…”

Section: Introductionmentioning

confidence: 99%

Modeling biological refractory periods and synaptic depression in an artificial neuron

Hamel¹,

Labib²

2019

Biomed. Phys. Eng. Express

View full text Add to dashboard Cite

In this paper, we increase the classifying potential of an artificial neuron, the quantron, by considering biologically inspired modifications. Synaptic depression, representing the decreasing phenomenon observed over time in an active neuron is modeled first, resulting in highly nonlinear and convex classification boundaries. Secondly, refractory periods during which receptors are partially inactive are incorporated into the model. Again, convex discriminant functions are obtained. Combining both of these improvements allows the quantron to generate even more complex patterns, resulting in a more versatile classifier.

show abstract

Section: Introductionmentioning

confidence: 99%

Modeling biological refractory periods and synaptic depression in an artificial neuron

Hamel¹,

Labib²

2019

Biomed. Phys. Eng. Express

View full text Add to dashboard Cite

show abstract

“…Training neural networks is by minimization a cost function defined using the output of the network and measurements from the modeled system (Zhang and Suganthan, 2016). Classical training approaches use derivatives of the cost function to update the weights of the neural network (Gori and Tesi, 1992;Montingy, 2014). Unfortunately, classical approaches do not guarantee finding the global minimum of the optimization problem and are often trapped in a local minimum.…”

Section: Introductionmentioning

confidence: 99%