blogusz / Neural-networks-and-Deep-Learning

Tasks carried out as part of the Neural networks and Deep Learning course.

Lab1 - Basic matrix operations in NumPy. Visualization of neural network weights.

1. Implementation of the sigmoid activation function.
2. Implementation of feed-forward operation in a single-layer neural network.
3. Implementation of the tiles function for visualizing weights in a neural network layer.

Lab2 - Data visualization and classification with scikit-learn and matplotlib.

Lab3 - Contrastive Divergence Algorithm.

1. Implementation of the reconstruction error in Restricted Boltzmann Machines.
2. Implementation of the CD-k algorithm.

Lab4 - Persistent Contrastive Divergence Algorithm.

1. Implementation of the PCD algorithm.

Lab5 - Momentum Method. Deep Belief Networks.

1. Implementation of the momentum method in CD-k.
2. Implementation of training and sampling in Deep Belief Networks.

Lab6 - Backpropagation Algorithm.

1. Implementation of the backpropagation algorithm.

Lab7 - Regularization and pre-training in deep networks.

1. Implementation of L1 and L2 costs in RBM.
2. Pre-training of deep MLP.

Lab8 - ReLU activation function. Regularization with weight vector norm.

1. Implementation of a limit on the weight vector norm.
2. Deep neural networks with ReLU activation function.

Lab9 - Dropout.

1. Implementation of Dropout regularization.

Lab10 - Autoassociative networks.

1. RBM with Gaussian units.
2. Implementation of an autoassociative network with a linear encoding layer.

Lab11 - Nesterov method and data visualization with autoassociative networks.

1. Implementation of the Nesterov method in an autoassociative network.
2. Visualization of the MNIST dataset using an autoassociative network.

Lab12 - Convolutional Neural Networks.

1. Implementation of a convolutional neural network.

Lab13 - Negative Sampling Algorithm.

1. Implementation of the Negative Sampling algorithm.