Name
Abstract | ||
|---|---|---|
Single neuron dynamics is derived in a coherent interacting mean field with all other neuron dynamics coupled through the Hebbian learning of synaptic weight updates. This exact one-body equation is derived because of the product nature of Hebbian learning rule. It provides the diagnoses needed for the collective interaction of artificial neural networks (ANN). In case of a chaotic neurodynamics generated with the N-shaped sigmoidal function, we can use the meanfield equation to demonstrate any single neuron behavior in the network, e.g. contrast reversal, limited cycles, chaos etc., due to a changing habituation threshold value. In case of negative threshold values, which might be viewed as an external reward to encourage neurons firing with higher rates, an image block quantization compression artifact has been quickly overcome with such a chaotic ANN and the result is quantified with a fractal dimensionality analysis of the image quality |
Year | DOI | Venue |
|---|---|---|
1997 | 10.1109/ICNN.1997.616098 | Neural Networks,1997., International Conference |
Keywords | Field | DocType |
hebbian learning,chaos,data compression,image coding,neural nets,hebbian learning rule,n-shaped sigmoidal function,changing chaos threshold value,changing habituation threshold value,chaotic neurodynamics,coherent interacting mean field,contrast reversal,coupled neuron dynamics,exact one-body equation,external reward,fractal dimensionality analysis,image block quantization compression artifact,image quality,limited cycles,meanfield chaos dynamics,single neuron dynamics,synaptic weight updates,artificial neural network,fractals,hebbian theory,quantization,artificial neural networks,image analysis,dimensional analysis,limit cycle,mean field | Statistical physics,Fractal,Curse of dimensionality,Hebbian theory,Artificial intelligence,Artificial neural network,Quantization (signal processing),Chaotic,Synaptic weight,Mathematics,Sigmoid function | Conference |
Volume | ISBN | Citations |
2 | 0-7803-4122-8 | 2 |
PageRank | References | Authors |
0.51 | 0 | 2 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Harold Szu | 1 | 149 | 38.33 |
| c c hsu | 2 | 4 | 0.97 |