Simple convolutional neural network for digit recognition

Assignment for “Knowledge technology”, course in the Master in Methodology for behavioral sciences.

Description

The purpose of the the project wasn’t to develop a sophisticated architecture (see the image below to get a sense of what kind of network we built) or to get highest accuracy in a test set. Rather we had to show that we had understood the fundamental concepts of convolutional networks.

The most interesting thing about this project was the fact that I had to build my own digit data set and then use it as test set. If you want to know more about it you can check the report I made. The supplementary material can be found on my github profile. I also made a few ppt slides to visually represent the net architecture. You can view the pdf here.