- Implemented multi-layer neural networks for image reconstruction
- Applied backpropagation to memorize images based on 2-dimensional (x, y) coordinates
- Explored various input feature mapping techniques following recommendations from a relevant paper
- Developed forward and backward passes for the network
- Trained a four-layer network using both Stochastic Gradient Descent (SGD) and Adam optimizers
- Ensured network weights were updated to minimize Mean Square Error (MSE) loss between original and reconstructed images
- Managed hyperparameters, including hidden layer size, learning rate, and activation function choices.
- Documented project details and findings in the ’neural_network.ipnyb’ notebook