Machine Learning for Robotics

This course provides a broad overview of machine learning techniques and their practical application in robotics.

Intro to Machine Learning and python

• Machine Learning in Robotics

• Understanding the goal of machine learning, fundamental problems, high-level overview. Visualizing your data

• Supervised vs Unsupervised vs Reinforcement learning

• Model-driven vs Data-driven: White box, Gray box, Black box

• Prior knowledge vs Unstructured. Feature extraction, linear regression + Deep Learning.

• Interpretability

Hands-on Machine Learning

• Example machine learning project in Robotics

• Binary classification, decision boundaries, using logistic regression/SVM (one parameter)

• How to perform a classification experiment. Dataset splitting, learning curves, metrics, comparing results

Regression and Data collection

• Regression and Data Collection in Robotics

• Regression methods, least squares fitting

• Overfitting, cross validation, regularization

• Collecting (noisy) data, labelling data, outliers

• High-dimensional data, data augmentation

• Hyper-parameter optimization (grid search vs random search)

Classification

• Classification in Robotics

• Parametric vs Non-parametric classifiers

• Logistic regression

• Decision tree, forest

• Bayesian classification: Bayes' rule, naive Bayes, Gaussian Mixture Model

• k-nearest neighbour

• SVM, kernel-SVM, dual problem

• Multi-class classification, metrics (confusion matrix), class imbalance

Unsupervised Learning

• Unsupervised Learning in Robotics

• Clustering: K-means, Gaussian Mixture Model, DBSCAN

• Dimensionality reduction: PCA, Local Linear Embedding (LLE)

Neural Networks

• Neural Networks in Robotics

• Multi-Layer Perceptron, gradient descent

• Neural Networks, and deep learning

• Vanishing gradient problem, DropOut, Optimizers

• Recurrent neural networks

• Convolutional neural networks

• AutoEncoder

• Data augmentation

Advanced machine learning

• Reinforcement learning

• Robot Learning

Final assignment (30%)

• Pairwise collaboration

• Late submission: -1 grade point per day

• No resit

• Deadline: Will be released after the 6th practicum, giving the students approx. two weeks to read the assignment, collect the necessary data, perform their experiments, and write their report

Exam (70%)

• Individual

• Written closed-book exam

• Exam in Q1, resit in Q2

Knock-out criteria

• At least 4 practicum grades need to be a pass (of current or previous academic year)

• Minimum grade for final assignment is a 5 (of current or previous academic year)

• Minimum grade for the exam is a 5

• Final grade minimum is a 6