Repository for the individual assignments (1 and 2) from the Deep Learning course at the Vrije Universiteit taught by Peter Bloem, Michael Cochez and Shujian Yu.

In this assignment we worked both with syntetic data and the MNIST dataset, we had to:

• Implement a small neural network from scratch, using only Python's Math package, reporting our math derivation for the derivatives.
• Vectorize the forward and backward operatrions for the above using the numpy package.
• Allow for processing in batches (batched gradient descent).
• Test our final model on the MNIST dataset and report the results

Grade: 9/10

In this assignment we focused on how an autograd library is implemented through the vugrad educational library and PyTorch. More specifically we had to:

• Provide the mathematical derivation of the vectorized forward and backward pass for different matrix operations.
• Explain and implement part of the necessary steps and components of an autograd library software architecture.
• Fine-tune the parameters of the network and test it on the MNIST-dataset
• Implent another NN classifier with PyTorch, fine-tune and test it on the image dataset CIFAR10.

Grade: 8/10