Abstract:Based on the network structure and training methods of extreme learning machines, extreme learning machine combining hidden-layer feature weighting and batch training (ELM-WB) is proposed to make full use of representation-level features for object images and human action videos classification. To solve the problem of insufficient fusion of multiple representation-level features in most classification methods, a double hidden layer structure in which the input layer and the second hidden layer are directly connected is designed. A loop training method of weighting coefficients and output weights is proposed based on the advantages of this structure. The proposed network structure and training method are combined to construct an extreme learning machine combining hidden-layer feature weighting (ELM-W), which can effectively fuse representation-level features to enhance the classification ability of representation-level features. On this basis, the principle of online sequential ELM (OS-ELM) is introduced to update the loop training formula of the two weights to reduce the memory consumption during the operation of the overall algorithm. ELM-WB is proposed by combining the loop training formula of two weight matrices with batch training. In order to test the feasibility of ELM-WB, experiments are conducted on Caltech 101, MSRC, UCF11 and UCF101 databases. Experimental results prove that the proposed ELM-WB can improve classification accuracy by fusing representation-level features. At the same time, ELM-WB can be used to perform classification tasks on databases of any size in a general-purpose computer without specific hardware.

Deep Convolutional Representations and Kernel Extreme Learning Machines for Image Classification

Multiple-kernel-learning-based Extreme Learning Machine for Classification Design

Database Systems for Advanced Applications

Extreme Learning Machine Combining Hidden-Layer Feature Weighting and Batch Training for Classification

Research on Image Classification Method of Features of Combinatorial Convolution

Extreme Learning Machine with Kernel Model Based on Deep Learning

ELMAENet: A Simple, Effective and Fast Deep Architecture for Image Classification

A Novel Bio-inspired Image Recognition Network with Extreme Learning Machine

A Deep Neural Network Based on ELM for Semi-supervised Learning of Image Classification

Classification of Hyperspectral Imagery Using a New Fully Convolutional Neural Network

A Novel Biologically Inspired ELM-based Network for Image Recognition

Densely connected convolutional extreme learning machine for hyperspectral image classification

Learning Deep Classifiers With Deep Features

A Spectral-Spatial Domain-Specific Convolutional Deep Extreme Learning Machine for Supervised Hyperspectral Image Classification.

Classifying Network Data with Deep Kernel Machines

Extreme Learning Machine and Adaptive Sparse Representation for Image Classification

Supervised Classification of Hyperspectral Images Using Local-Receptive-fields-based Kernel Extreme Learning Machine.

A Research on Image Recognition and Classification Based on Traditional Machine Learning and Deep Learning

Kernelized Support Vector Machine with Deep Learning: an Efficient Approach for Extreme Multiclass Dataset

Learning Kernel for Conditional Moment-Matching Discrepancy-Based Image Classification

Exploiting Convolutional Neural Networks with Deeply Local Description for Remote Sensing Image Classification.