Image recognition method based on adaptive matrix iterative extreme learning machine

Abstract

The invention discloses an image recognition method based on an adaptive matrix iterative extreme learning machine. The method comprises the following steps: combining a network structure of a traditional extreme learning machine with a self-adaptive matrix iteration mode, and performing convergence by using a convergence factor self-adaptive matrix iteration mode to obtain an output weight matrix when solving the output weight matrix; therefore, while the characteristics of simple network structure, random parameter generation and the like of a traditional extreme learning machine are kept; the method has better training precision in the aspect of image recognition, occupies less computing resources and consumes less time. A new thought and a new approach are provided for improvement and optimization of a machine learning algorithm and image recognition.

Description

technical field [0001] The invention belongs to the technical field of image recognition based on machine learning, and in particular relates to an image recognition method based on an adaptive matrix iterative extreme learning machine. Background technique [0002] With the breakthrough development of artificial intelligence, machine learning, as a mainstream method to solve artificial intelligence problems, is constantly innovating and improving. As an important field of artificial intelligence, image recognition technology can accomplish tasks that general sensors cannot. Image recognition technology processes image data based on image features, and uses algorithms to build models. The quality of the algorithm will determine the effect of image recognition. [0003] Extreme Learning Machine (Extreme Learning Machine, ELM) is a feedforward neural network algorithm with a single hidden layer. Due to its simple structure, input weights and deviations can be randomly generat...

AI Technical Summary

Problems solved by technology

Although the traditional ELM algorithm has the advantages of simple network structure, fewer setting parameters, and faster training speed compared to multi-layer neural networks, due to its single-layer network structure and parameter randomness, the accuracy of the learning model it trains is limited. has great instability

At the same time, the method of solving the Moore-Penrose generalized inverse used to solve the output weight matrix also has certain defects. This method will lead to long calculation time and low image recognition accuracy in some cases.

Images