Abstract: Binary neural network (BNN) is a specialized form of neural networks designed to adapt to re-source-constrained environments. I's core characteristic is that both the network weights and activationfunction outputs are binarized, i. e. , converted entirely to the binary form of O or 1. BNN can reducethe model's storage space occupation and computing resource consumption, thus being widely applied invarious mobile devices and embedded systems. To enhance the computational efficiency of BNN and re-duce their energy consumption, this paper analyzes the core concepts of neural network binarization andkey points of framework construction, and proposes a method for building a BNN learning framework u-sing Matlab. This provides academic references and technical support for BNN to belter utilize limited re-sources and for deploying and running deep learning models in various environments

Key words: Matlab, Binarytraining functions Neural Network(BNN), learning framework, network parameters, training functions