Statistical Properties of Chaos Associative Memory
スポンサーリンク
概要
- 論文の詳細を見る
In this report we shall propose a chaos dynamic memory model applied to the chaotic autoassociation memory. The present artificial neuron model is properly characterized in terms of a time-dependent sinusoidal activation function to involve a transient chaotic dynamics as well as the energy steepest descent strategy. It is elucidated that the present neural network has a remarkable retrieval ability beyond the conventional models with such a monotonous activation function as sigmoidal one. This advantage is found to result from the property of the analogue periodic mapping accompanied with a chaotic behaviour of the neurons as well as the symmetry of the dynamics equation.
- 社団法人電子情報通信学会の論文
- 2001-03-10
著者
関連論文
- Entropy Based Associative Memory(Selected Papers from the 18th Workshop on Circuits and Systems in Karuizawa)
- A generalised entropy based associative memory (非線形問題)
- Statistical Properties of Chaos Associative Memory
- NLP2000-46 / NC2000-40 A Chaos Memory Retrieval with a Skew-Tent Activation Function
- Statistical Properties of Chaos Associative Memory
- A Study of Chaos Synergetic Neural Network
- On the memory capacity and invariant measure of chaos associative models (非線形問題)
- On the memory capacity and invariant measure of chaos associative models (回路とシステム)
- Chaos Associative Memory with a Periodic Activation Function
- NLP2000-46 / NC2000-40 A Chaos Memory Retrieval with a Skew-Tent Activation Function