Structure optimization method of reserve pool network based on excitability and inhibition STDP

Abstract

The invention provides a structure optimization method of a reserve pool network based on excitability and inhibition STDP. The reserve pool network is a network formed by mutually connecting excitatory neurons E and inhibitory neurons I, the network has four connections: E-E, E-I, I-E and I-I, the connection weight of the E-E is adjusted through excitatory STDP, and the connection weight of the I-E is adjusted through inhibitory STDP. According to the invention, through an unsupervised learning rule of interaction between excitability STDP and inhibition STDP, excitability and inhibition balance is realized; a heterogeneous network structure which is consistent with the biological neural network in characteristics and has long-tail distribution characteristics is formed, the network information capacity is large, and the network performance is optimal.

Description

technical field [0001] The invention belongs to the technical field of artificial neural networks, in particular to a method for optimizing the structure of a reserve pool network based on excitatory and inhibitory STDP. Background technique [0002] Echo state networks (ESNs) are a new type of neural network model. The echo state network belongs to a multi-layer feedforward neural network model in terms of macroscopic structure. Its structure is as follows: figure 1 As shown, the local recurrent neural network in the middle part of the figure is the reservoir computing module, which belongs to the key part of the echo state network. [0003] The reserve pool module is a recurrent neural network. The neurons of the reserve pool network are connected randomly and sparsely. The connection weight has a decisive impact on the system performance. In the prior art, how to set the connection weight of the reserve pool network The value is mainly adjusted by experience, and it need...

AI Technical Summary

Problems solved by technology

[0003] The reserve pool module is a recurrent neural network. The neurons of the reserve pool network are connected randomly and sparsely. The connection weight has a decisive impact on the system performance. In the prior art, how to set the connection weight of the reserve pool network The value is mainly adjusted by experience, and it needs experienced people to complete it, and it needs people with rich experience, because the amount of experience has a great impact on the quality of the network performance, and human subjectivity has a great influence, and the performance is not necessarily the same. best, and takes a lot of time

In addition, empirical data in some fields may be limited, and it is difficult to obtain a network structure with better performance

Images