In this paper we studied both HD and CNN for popular data set and compare the 2-approaches in terms of accuracy, computing complexity and overall performance. We optimized both HDC and CNN approaches independently and then analyzed the results to guide for the selection based on target need. MNIST dataset was used in this work for HDC-based classification. The code is inspired and modefied from the paper " Binary-Hyperdimensional-Computing-Trade-offs-in-Choice-of-Density-and-Mapping" to match the targeted application.
The folder HDC contains the main MNIST_encoding.mat, whih are the main program to simulate with all required dataset and functions. For the execution, uncomment the required vector dimention and the determine the item memory seeds.
The folder CNN, has the CNN model which used for this study.