Minimizing Time Complexity by Using Modified Strassen’s Matrix Multiplication Algorithm on Convolutional Layers

Abstract

Digital image processing refers to the processing of digital images by a computer. An image is composed of a finite number of elements each having a particular location and value. For image recognition in python, there are many layers called convolutional layers. Each of these layers requires multiplication with one another to generate a result. This is generally done by methods like Linear Discriminant Analysis (LDA) and Principal Component Analysis (PCA), which were time-consuming. Using our modified Strassen's matrix multiplication algorithm on the image data yields a better result as it can provide the same accuracy by training more images in lesser time. By using our modified algorithm, we were able to reduce the time by 37.408 %. This study helps in providing a better approach to image recognition than the general approach maintaining the same accuracy rate but reducing the time to achieve it.