Implementation of a feed-forward neural network in C++ from scratch. This is a task of the Neural Networks course of Masaryk University (Brno, Czech Republic). It reaches at least 88% of correct test predictions (overall accuracy). It uses L2 regularization. In order to be faster it's been implemented using OpenMP (16 threads).
The dataset is Fashion MNIST [0], a modern version of a well-known MNIST [1]. Fashion-MNIST is a dataset of Zalando's article images ‒ consisting of a training set of 60,000 examples and a test set of 10,000 examples. Each example is a 28x28 grayscale image, associated with a label from 10 classes. There are four data files — two data files as input vectors and two data files with a list of expected predictions.
The implementation exports vectors of test predictions. Such number on i-th line represents predicted class index (there are classes 0 - 9 for Fashion MNIST) for i-th input vector. Exported files are called actualTestPredictions.
These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.
You can clone the project using:
git clone https://github.com/guillempla/Fashion-MNIST-Neural-NetworkYou can compile, execute and export everything with:
./RUNIt will read data, train the model, and generate the predictions file called actualPredictions.
Note: You can remove "module add gcc-10.2" from "RUN". It's only necessary for tesing on AISA computer
- Guillem Pla Bertran - guillempla@protonmail.com