Instructor: Dr. Amir Jafari | GitHub: https://github.com/amir-jafari
The main focus of this course will be the implementation of deep learning techniques on GPUs. Four key deep learning architectures will be covered. Multilayer Perceptrons, Convolution Networks, Deep Belief Networks and Long Short Term Memory are the main four deep network architecture. Sometime will be spent on the background of each network, but the primary focus will be on implementation. In addition to discussing the four network architectures, the course will concentrate on four of the most popular deep learning frameworks: Torch, Caffe, Theano and Tensorflow. The strategy will be to present a deep network architecture, and then describe how that network can be trained and analyzed within a particular framework. Each network will be trained in a different framework.
- implement the machine learning algorithms on CPU and GPU.
- use four deep learning architectures (Multilayer Perceptrons, Convolution Networks, Deep Belief Networks and Long Short Term Memory).
- train and test four deep learning network architectures (MLP, CNN, DBN, LTSM).
- use the four popular software in the deep learning area Caffe. Torch, Theano and Tensorflow.
- train and analyse within a particular framework.
- MiniProject I - Training Multilayer Preceptron with Torch - MNIST Dataset
- MiniProject II - Training Convolution Network with Caffe - MNIST Dataset
- Final Project - Plant Seedlings Classification - Kaggle Competition