The tentative syllabus is given below. Please be informed that the timeline and lecture topics are subject to change. For each week, we will regularly update the course material, assignments, and suggested readings.
Course overview and logistics
* The brief history of machine learning and Humanoid robotics
* Introduction to FOSS robot simulators
* Math and programming resources for multimodal machine learning
DL Book, Part 1
Intro to machine learning and robotic sensors
* ML terminology: targets, features, priors, models, hyperparameters, etc.
* Robotic sensors: camera, depth, position encoders, ultrasound, infrared, and lasers
* Fantastic machiine learning and robotics datasets and where to find them
DL Book, Part 1
Intro to neural networks (NNs), multimodal fusion techniques and software stack
* Challenges in multimodal learning: data representation and fusion
* NN basics: single and multilayer perceptron, parameter initialization, gradient descent, backpropagation, activation and loss functions, etc.
DL Book, Ch 6
The pepper robot
Intro to Convolutional Neural Networks (CNNs)
* CNNs concepts: convolution, pooling, stride, padding, etc.
* CNN architectures: Neocognitron, LeNet, AlexNet, VGGNet, ResNet, etc.
DL Book, Ch 9
CNN visualizer Feature visualization
Reinforcement learning basics and deep reinforcement learning
* RL concepts: Markov decision processes, exploration/exploitation, convergence requirements, model-free methods (e.g., Q learning, SARSA), etc.
* Challenges in RL: credit assignment, exploration/exploitation trade-off, etc.
* Deep Q Learning variations: DQN, Double DQN, Dueling DQN, DQN experience replay, etc.
RL Book, Ch 3, Ch 6
Key papers in RL
OpenAI RL resources
Intro to Humanoid robot simulators and demos on the Pepper robot
* Installation of pyBullet/qiBullet simulators
* Demos with the Pepper and Nao robots: reading senosory, kinematics, random sensorymotor babbling, etc.
Intro to unsupervised learning
* Applications of unsupervised learning: vector quatization, clustering, feature extraction.
* Learning methods: K-Means, competitive learning and self-organizing maps
* Unsupervised learning robot architectures: PyEra and SOIMA
Elements of statistical learning Ch 14.3, 14.4
|W8-W14||The presentation dates are tentative and it will be finalized after the 2nd assignment submission.|
|The template was modified by Murat Kirtay by using
|© Conference Organizers.|