The purpose of this paper is to propose the circuit for autonomous determination of optimal NN (neural network) structure for the applied problems and to implement the circuit on an FPGA (field programmable gate array) to test its basic functions. Many research activities to determine optimal structure of the NN have been done so far, however, no NN LSI (neuro-chip), which can execute NN functions quickly, can determine appropriate network structure autonomously, i.e. its network structure is determined by using software running on a PC (personal computer) or a WS (work station). In this paper, to determine the optimal network structure for the neuro-chip autonomously (i.e. without a PC or a WS), we propose hardware implementation of GA (genetic algorithm) operations, which can be embedded on the neuro-chip.
It is difficult to realize the context dependency of cognition in real world environment using conventional pattern recognition technique. We have proposed memory model PATON that realizes such cognition based on attention. To apply the PATON to real environment, we introduce a continuous-time dynamics, a mechanism of event driven attention generator, and an autoassociator network with constraints for encoding of external input in various modality. Then it is shown that the ability of multi-modal information processing of PATON is useful for continuous spoken word recognition task with visual input.
To understand the inhibitory network in the cortex, we have started to clarify several features of GABAergic neurons in the frontal cortex of rats. These are; (1) the physiological, chemical and morphological characteristics of cortical GABAergic neurons to identify the subtypes, (2) the synaptic connections among these neuronal subtypes and (3) the synaptic properties and the chemical modulation of each subtype. An experimental approach in which multiple properties of a GABA neuron are determined for the same cell, yields a set of data which allows a more comprehensive classification of the neuron as an element of a local neuronal circuit.