Convolution calculation method and device based on sparse neural network, and electronic equipment

The invention relates to the technical field of neural networks, in particular to a convolution calculation method and device based on a sparse neural network and electronic equipment, and the method comprises the steps: determining a splitting rule of a filter array according to the convolution operation specification in the neural network, and splitting the filter array to obtain a plurality of groups of sub-filter arrays; respectively carrying out sparse processing on each sub-filter array obtained after splitting to enable the weight in a target weight column in each sub-filter array to be zero; aiming at weights in the effective weight columns contained in the sub-filter arrays, sequentially storing the weights in a weight memory according to the sequence of the effective weight columns; in the convolution calculation process, the weights in the effective weight columns contained in the sub-filter arrays in the weight memory are sequentially read, and convolution calculation is carried out through the re