A convolution reserve pool optimization method based on an evolutionary chaos edge

Abstract

The invention relates to a convolution reserve pool optimization method based on an evolutionary chaos edge, which is used for improving the computing power of a reserve pool and comprises the following steps of: 1) dividing a time sequence Y to be processed and a related input variable X into a training data set and a prediction data set; 2) randomly initializing an input connection matrix and anlth-order delay connection matrix of the storage pool, and setting a maximum iteration frequency and a test error threshold; 3) initializing hyper-parameters; 4) setting a parameter search space; 5)inputting the training data set into a convolutional echo state network, and performing training to obtain an input connection estimation matrix; 6) taking hyperparameters of Particle swarm optimization-gravity search algorithm based on memory strategy and Levy random walk as the particle position, and carrying out chaos edge reserve pool optimization, thereby obtaining the optimal hyper-parameterof the convolution reserve pool. Compared with the prior art, the method has the advantages of time dependence simulation, stable chaotic edge, high prediction classification performance and the like.

Description

technical field [0001] The invention relates to the field of information technology, in particular to a convolution reserve pool optimization method based on evolutionary chaos edge. Background technique [0002] Time series widely exist in the fields of finance, communication, meteorology, electric power, control, medical signal processing, etc., and the analysis and prediction based on time series has important practical engineering value. At present, the main method for processing time series is the recurrent neural network, but due to the problems of gradient fading and computational complexity in the recurrent neural network, its practical application is limited. [0003] Reserve pool computing, originally in the form of echo state networks and liquid state machines respectively, has grown into two separate subfields and attracted considerable attention over the past few years. It tries to build a dynamic reserve pool that operates on the "edge of chaos", with extremel...

AI Technical Summary

Problems solved by technology

From the perspective of biological evolution, variation evolution will lead to the complexity of the ecosystem, and will cause the system to be on the edge of chaos or chaos. Therefore, biological variation evolution is a main demonstration of the edge of chaos, and it also shows that chaos (edge) is a A changing state rather than a static state, so automatic cell machines were first used for reserve pool optimization, but the rules used were not complex enough to meet the complexity requirements of the edge of chaos

[0006] Biological optimization algorithm is a commonly used parameter estimation method, but the existing method uses the current best fitness to update all particle position information, and it is easy to have local minimum points for complex parameter optimization on the edge of chaos

Images