[elm classification] data classification based on particle swarm optimization convolution neural network CNN combined with limit learning machine elm with matlab code

1 brief introduction

Convolutional neural network is a good feature extractor , But it is not the best classifier , And the extreme learning machine can classify well , But can't learn complex characteristics , According to the advantages and disadvantages of both , Combine them , A new data classification method is proposed .

Considering that convolution neural network can extract optimal classification features , And the training speed of extreme learning machine is fast 、 training Practice fine degree high , Ben writing carry Out An algorithm that combines the two , And applied to data classification . The characteristic of this method is ： First, the training samples are used to train the convolutional neural network , After training , Other layer parameters of convolution network remain unchanged , And replace the last layer with the limit learn xi machine , Then the parameters of the extreme learning machine are trained , It is equivalent to taking the features extracted by convolution network as the input of limit learning machine , It combines the advantages of convolution network and limit learning machine . Although the weight sharing structure of convolutional neural network reduces the network parameters , But because of its complexity Of many layer junction structure , There are still a lot of parameters to train . When the number of training samples is small , The recognition effect is often not very good . On this basis, this paper proposes particle swarm optimization algorithm to optimize parameters , So as to improve the recognition accuracy .

2 Part of the code

function [XV]=FnBringtoRangeLowUpIPSODL(XV,XV_lower, XV_upper)​% dioperasikan untuk menandai yang nilainya kurang dari batas lowerXV_minus_XV_lower=XV-XV_lower;​% dioperasikan untuk menandai yang nilainya lebih dari batas upperXV_minus_XV_upper=XV-XV_upper;​% mendapatkan index yang nilainya kurang dari batas bawahidx_KurangDariLower=find(XV_minus_XV_lower<0);if(isempty(idx_KurangDariLower)) else    XV(idx_KurangDariLower)=XV_lower(idx_KurangDariLower);end​% mendapatkan index yang nilainya lebih dari batas atasidx_LebihDariUpper=find(XV_minus_XV_upper>0);if(isempty(idx_LebihDariUpper)) else    XV(idx_LebihDariUpper)=XV_upper(idx_LebihDariUpper);end

3 Simulation results

4 reference

[1] Zhoujiayin . Research on image classification algorithm based on convolution neural network and limit learning machine [D]. Guangdong University of Technology .

About bloggers ： Good at intelligent optimization algorithms 、 Neural networks predict 、 signal processing 、 Cellular automata 、 The image processing 、 Path planning 、 UAV and other fields Matlab Simulation , relevant matlab Code problems can be exchanged by private letter .

Some theories cite network literature , If there is infringement, contact the blogger to delete .