CS 598 Deep Learning, UIUC Machine Problems
Problem 1: Implementation of fully connected neural network from scratch using numpy
Implement and train a neural network from scratch in Python for the MNIST dataset (no PyTorch). The neural network should be trained on the Training Set using stochastic gradient descent. Target accuracy on the test set: 97-98%
Problem 2: Implementation of Convolution Neural Network from scratch using numpy
Implement and train a convolution neural network from scratch in Python for the MNIST dataset (no PyTorch). The convolution network should have a single hidden layer with multiple channels. Target accuracy on the test set: 94%
Problem 3: Implementation of Deep CNN for CIFAR 10
Train a deep convolution network on a GPU with PyTorch for the CIFAR10 dataset. The convolution network should use dropout, trained with RMSprop or ADAM, and data augmentation. Also, compare dropout test accuracy using the heuristic prediction rule and Monte Carlo simulation.
Problem 4: Implementation of Deep Resnet on CIFAR 100
Implement a deep residual neural network for CIFAR100 using your own Resnet and pretrained Resnet implemented in Pytorch. Target test set accuracy: 60% for own Resnet implementation and 70% for pretrained Resnet implementation.
Problem 5: Image Similarity with Deep Ranking on Tiny Imagenet
Problem 6: Implementation of GAN with Wasserstien Gradient Penalty (Wasserstien GAN) and Auxilliary Classifier (ACGAN) on CIFAR 10