Neural network implementation - university project.

Task: Classify FashionMNIST images.

Consists of:

• MLP architecture + various activations with hardcoded derivations (no autograd, sorry ¯/_(ツ)_/¯)
• Adam and SGD optimizers
• Weight initializations
• Multicore processing with OpenMP
• Train-val-test split

Missing: batch shuffling, early stopping, ...

Achieves 87% accuracy.

Disclaimer: All subsequent submissions of the recurring project at the respective uni are automatically checked for plagiarism against this repository source codes. Keep that in mind.