omedusyo/deep-learning
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
Solo School Project for a Neural Networks Class implementing backprop/MLP on a MNIST dataset
1 commit 1 branch 0 tags 65 MiB
Find a file
Exact
Yura Dupyn fa8eae3146 Initial commit
2026-03-20 11:48:07 +01:00
data Initial commit 2026-03-20 11:48:07 +01:00
src Initial commit 2026-03-20 11:48:07 +01:00
.gitignore Initial commit 2026-03-20 11:48:07 +01:00
Cargo.lock Initial commit 2026-03-20 11:48:07 +01:00
Cargo.toml Initial commit 2026-03-20 11:48:07 +01:00
OVERVIEW.png Initial commit 2026-03-20 11:48:07 +01:00
README.md Initial commit 2026-03-20 11:48:07 +01:00
tmp_repl.txt Initial commit 2026-03-20 11:48:07 +01:00

README.md

Rust Deep Learning

School Project I did for a Neural Networks Class at the end of 2023:

  • Implements a neural net (backpropagation / multilayer perceptron) in Rust
  • Constraint: Can't use any Linear Algebra libraries or frameworks
  • Training/Dataset: Fashion-MNIST dataset (achieves about 91% accuracy in less than 10 min of training)

See OVERVIEW.png for the underlying math I came up with to do backprop and organize memory.

Dataset

Fashion MNIST (https://arxiv.org/pdf/1708.07747.pdf). Dataset of 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. The dataset is in CSV format:

  • fashion_mnist_train_vectors.csv - training input vectors
  • fashion_mnist_test_vectors.csv - testing input vectors
  • fashion_mnist_train_labels.csv - training labels
  • fashion_mnist_test_labels.csv - testing labels