Width learning method based on a minimum P norm

Abstract

The invention discloses a width learning method based on a minimum P norm. The width learning method comprises the steps of 1, obtaining training input data and training output data; 2, generating a hidden layer node output matrix through randomly generated weights and offsets by adopting the same mode as a width learning system; step 3, taking the P norm of the error as a cost function, and solving an output weight by combining a fixed point iteration strategy; and 4, estimating the output corresponding to the test input by using the trained model parameters. Due to the fact that the BLS cannot effectively complete regression and classification tasks in the presence of complex noise or abnormal interference, the invention provides the width learning method based on the minimum P norm. According to the method, the characteristic that the P norm of the error can well cope with different noise interferences is utilized, so that the regression and classification tasks can still be smoothly completed under the condition that complex noises or abnormal interferences exist, and the method has important research significance and wide application value.

Description

【Technical field】 [0001] The invention relates to a width learning method based on the minimum P norm. 【Background technique】 [0002] The deep neural network learning model has been successfully applied to the modeling of many regression and classification problems. Typical deep neural network learning models include Deep Boltzmann Machines (DBM), Deep Belief Networks (DBN), Convolutional Neural Networks (CNN), etc. In order to effectively establish a network model, these deep learning methods need to continuously adjust the number of layers of the neural network and the number of nodes required by each layer of the network, and then iteratively determine the connection weight between each layer and the layer. When the amount of data is very large, this adjustment is very time-consuming, which brings difficulties to practical applications. [0003] In order to solve this problem, Chen Junlong of the University of Macau proposed a breadth learning method based on the Rando...

AI Technical Summary

Problems solved by technology

Since it is difficult to effectively eliminate the negative effects of non-Gaussian noise or outliers due to the second-order statistical characteristics of minimizing errors, the learn

Images