Assignments

Assignment 1

Complete the Deep Learning main tutorial: http://deeplearning.net/tutorial/
You do not need to hand in anything. However, make sure to go through the tutorial and have a good understanding of the material. You will need it for the next assignments. It may also be a valuable tool if you choose to do a project.

Due: February 12th (midnight)

Other resources
Learning from Data. Caltech class on Machine Learning:  https://work.caltech.edu/telecourse.html
More deep learning tutorials and readings: http://www.kdnuggets.com/2014/05/learn-deep-learning-courses-tutorials-overviews.html

Assignment 2

Classify MNIST data using Convolutional Neural Networks.
Read the following resources, meant to help with this task. The first two introduce convolutional neural networks and ML techniques (applied to the MNIST dataset) for beginners. The third one is a tutorial on creating a multi-layer perceptron. The last resource is an implementation for MNIST recognition on a two layer neural network. For this assignment, you could use packages (such as Caffe, Theano) or existing convolutional neural networks (such as LeNet), as long as you understand what you are doing.

Due: February 26th (midnight)

Resources:                                                                                    
Convolutional Neural Networks: https://www.tensorflow.org/versions/r0.7/tutorials/deep_cnn/index.html

MNIST for ML Beginners: https://www.tensorflow.org/versions/master/tutorials/mnist/beginners/index.html

Multi-layer perceptron tutorial: http://deeplearning.net/tutorial/mlp.html

MNIST recognition for 2-layer-perceptrong: http://davidstutz.de/recognizing-handwritten-digits-mnist-dataset-twolayer-perceptron/

Assignment 3

Generate MNIST data using Recurrent Neural Networks.
In Assignment 2, we worked on classification. Now, we focus on generating MNIST digits. Try to implement this as done in this paper: http://arxiv.org/pdf/1502.04623.pdf. The assignment is 2 weeks long, however you will have to give an intermediate progress report after the first week.

Due: March 11th (midnight)