Course

CS-456: Deep reinforcement learning

Lectures in this course (96)

Mini-Batches in On- and Off-Policy Deep Reinforcement Learning

Explains the significance of mini-batches in Deep Reinforcement Learning and the differences between on-policy and off-policy methods.

Tricks of the Trade in Deep Learning: Aims

Covers practical questions and aims in deep learning, including neuron types, network architecture, optimization, and weight initialization.

Proximal Policy Optimization for Continuous Control

Explores Proximal Policy Optimization for enhancing stability and efficiency in continuous control with deep reinforcement learning.

Deep Deterministic Policy Gradient for Continuous Control

Presents the Deep Deterministic Policy Gradient algorithm for training neural networks in continuous action spaces efficiently.

Theory of Bagging

Explores the theory of bagging, demonstrating how it improves model performance and the importance of uncorrelated data for its success.

Bagging: Regularization Method in Deep Learning

Explores bagging as a regularization method in deep learning, training multiple model variants on different data subsets to improve generalization.

Monte Carlo Tree Search and Alpha Zero

Explores Monte Carlo Tree Search and Alpha Zero in deep reinforcement learning.

MuZero: Planning and Learning Model

Covers MuZero, a model that learns to predict rewards and actions iteratively, achieving state-of-the-art performance in board games and Atari video games.

Model-Based Deep RL: Planning and VAST

Covers model-based reinforcement learning, planning, variational state tabulation, and efficient Q- and V-values updating.

Dropout: Tricks of the Trade

Explores Dropout as a regularization method in deep neural networks, emphasizing its practical implementation and effectiveness.

Data Augmentation: Deep Learning

Explores data augmentation as a key regularization method in deep learning, covering techniques like translations, rotations, and artistic style transfer.

Weight Initialization: Tricks of the Trade

Explores smart weight initialization in neural networks, emphasizing the importance of proper data normalization and random weight initialization.

Vanishing Gradient Problem: Deep Learning

Discusses the vanishing gradient problem in deep neural networks and its solutions.

Deep Learning: Backward Propagation and Vanishing Gradient

Delves into backward propagation in deep learning, addressing the vanishing gradient challenge and the need for effective hidden units.

Weight Update: Mean Input and Bias Problem

Discusses the mean input shift and bias problem in weight updates for neural networks, highlighting the importance of correct initialization to prevent gradient issues.

Batch Normalization: Why It Works

Explores the aim and process of batch normalization in deep neural networks, emphasizing its importance in stabilizing mean input and solving the vanishing gradient problem.

Variations of SARSA: Expected SARSA and Q Learning

Explores expected SARSA and Q learning, two variations of the SARSA algorithm.

TD Learning: Temporal Difference Learning

Covers Temporal Difference Learning, V-values, state-values, and TD methods in reinforcement learning.

Monte-Carlo Methods Quiz

Presents a quiz on estimating return variables in Monte-Carlo methods.

Monte-Carlo Methods for Reinforcement Learning

Explores Monte-Carlo methods for reinforcement learning, comparing them with TD-methods and emphasizing the efficiency of TD methods in propagating information.