Introduction to Deep Learning🔗
Abstract🔗
Deep Learning, a form of machine learning inspired by biological learning, has powered innovations in many fields, including Reinforcement Learning. While artificial neural networks, the core component of deep learning, have been used in artificial intelligence for decades, recent advances in the domain of deep learning such as GPU computation, convolutional layers, and data availability have led to breakthroughs in machine learning, computer vision, and many applications of deep learning such as protein folding. Deep Reinforcement Learning uses deep neural networks as a central component for many algorithms, such as Deep Q Networks and Soft Actor Critic; it is, therefore, important to understand deep neural networks for Reinforcement Learning. In this session, we will introduce the theoretical foundations of Deep Learning, namely backpropagation, stochastic gradient descent, and layer operations such as convolution. We will apply these concepts in exercises using the PyTorch library on supervised learning examples.