bogdanmagometa / nn

Manual Implementation of Deep Learning Models

Manual Implementation of Deep Learning algorithms

Project structure:

• nn/ - folder with implementation of automatic differentiation (AD) of Tensor operations, optimizers and parts of neural networks (linear layer, activation layers).
• tests/test_tensor.py - unit tests for automatic differentiation (AD) of Tensor operations.
• test_optimizers.ipynb - comparing four implemented optimizers.
• comparison_with_pytorch.ipynb - comparison of our optimizers with PyTorch ones.
• nn_models.ipynb - training MLPs using our implementation of AD.
• test_perceptron.ipynb - training perceptron using our implementation of AD to mimic AND gate.
• torch_models.ipynb - implementing the same models as in nn_models.ipynb (+ CNN) but in PyTorch.