Spatio-temporal associative memory and a high-order correlation neuronal network
✍ Scribed by Qian Gao; Zili Liu
- Publisher
- Elsevier Science
- Year
- 1988
- Tongue
- English
- Weight
- 50 KB
- Volume
- 1
- Category
- Article
- ISSN
- 0893-6080
No coin nor oath required. For personal study only.
✦ Synopsis
The elementary characteristic of a neural system is that its stable interactions are specified by temporal integration. Thus motivated spatio-temporal associative memory has been investigated in a model of neuronal network, the dyr~Amtc and the corresponding energy equations are also derived. With the introduction of temporal integrations, hetero-assoclative memory can be naturally re~lt=ed, yet the Hopfleld model is a second-order auto-associative one. Our high-order correlation network yields a larger storing capacity, and a learning rule is presented based on Kandel's work. Axon-axon synapses underlying learning and memory are the cellular basis of the conditioned reflex (Kandel et al.). They are thus proposed as the elementary units, which we call the Generalized Paviov-Kandel Units, in this work. Our study touches on some important issues in the neural mechanism of higher functions, and attempts to expintn the character that lower and higher avtm~l~ share the same cellular basis and connecting principles, but functionally differentiate on dynamical behavior. It is also suggested that a great Increasing number of synapses instead of neurons themselves is of physiological profound significance. And the essence of a network is its redundant correlated encoding.