In this repository the code implemented during the course Deep Learning in Data Science hold in Kungliga Tekniska Högskolan (KTH).
Assignment 1: Training and testing of a one layer network with multi-ple outputs to classify images from the CIFAR-10 dataset. The network will be trained using mini-batch gradient descent applied to a cost function that computes the cross-entropy loss of the classifier applied to the labelled training data and a L2 regularization term on the weight matrix.
Assignment 2: Training and testing of a two layer network with multiple outputs to classify images from the CIFAR-10 dataset. The network will be trained using mini-batch gradient descent applied to a cost function that computes the cross-entropy loss of the classifier applied to the labelled training data and a L2 regularization term on the weight matrix. Attention will be paid to how to search for good parameter settings for the network’s regularization term and the learning rate.
Assignment 3: Training and testing of a k-layer network with multiple outputs to classify images from the CIFAR-10 dataset. The network will be trained by minimizing a cost function, a weighted sum of the cross-entropy loss on the labelled training data and a L2 regularization of the weight matrices using mini-batch gradient descent.
Assignment 4: Training and testing a RNN to synthesize English text character by character. The text will be taken from the book The Goblet of Fireby J.K. Rowling. The variation of SGD used for the optimization will be AdaGrad.