Artificial Intelligence (30 hours including courses and lab works)
Artificial intelligence seeks to make high level cognitive functions calculable to allow the development of autonomous robots requiring little or no supervision. In this course we will discuss two of these main functions. We will first focus on the techniques resulting from AI work that allow a robot to make decisions to achieve a given objective autonomously. The field of AI, which is interested in this issue is called AI Planning. In a second step we will look at the techniques that allow a robot to learn to perceive its environment. The field of AI that is interested in this issue is called machine learning.
MACHINE and STATISTICAL LEARNING (30 hours, including courses and lab work)
ARTIFICIAL INTELLIGENCE (AI)
AI Planning is one of the fundamental sub-areas of Artificial Intelligence, concerned with algorithms that can generate strategies of action for arbitrary autonomous robots in arbitrary environments. The second part of the course will address so-called classical planning, where the actions and environment are assumed to be deterministic; this is a central area in planning, and has been the source of many influential ideas. It is also successfully applied in practice, as we will exemplify in the course. We will examine the technical core of the current research on solving this kind of problem. We will consider four different paradigms for automatically generating heuristic functions (lower bound solution cost estimators): critical paths, ignoring delete lists, abstractions, landmarks. Apart from understanding these techniques themselves, we will learn how to analyze, combine, and compare such estimators. We will furthermore consider optimality-preserving pruning techniques based on partial-order reduction, symmetries, and dominance pruning. The course contains many research results from the last decade, close to the current research frontier in planning.
AI-1: Introduction and Overview of Automated Planning?
AI-2: State Space Planning?
AI-3: Plan Space Planning?Lesson
AI-4: Planning-Graph Techniques
AI-5: Heuristics in Planning?
AI-6: Hierarchical Task Network Planning?
MACHINE LEARNING and Statistical LEARNING PRINCIPLES (ML)
This part of the course aims at providing the students with the basic methods and statistical/mathematical tools to understand the most classical approaches used in both supervised or unsupervised learning, as well as understanding their capabilities, characteristics and limitations.
ML-1: "Introduction to Machine Learning
- basic challenges: classification/regression, Supervised/unsupervised, cost functions, training error/testing
- examples of approaches WITH and WITHOUT models
- Notion of complexity: variance bias trade-off (over or under learning), scourge of the dimension
ML-2: Generative Approaches :
- Discriminant analysis (linear/quadratic, regularization, size reduction)
- Naive Bayesian (parametric and non-parametric
- examples given MNIST, ...
ML-3: Discriminatory approaches :
- generalized linear statistical models for regression and classification (logistic regression)
- ridge regulation, lasso (parsimony bet)
- Complexity and validation (cross-validation) of models
- Data transformation (kernel move)
- Examples given MNIST, ...
ML-4: "DATA REPRESENTATION" (Unsupervised).
- BCP, standardization, small visualization
ML-5. "(Convolutional) Neural Nets."
- Notion of neuronal learning, gradient stoch. GRADIENT, NN
- Deep, convolutional aspects... motivations and examples (scikit-learn, kheras/tensorflow,...)
ML-6. Random trees and forests
- decision tree, random forests
- Kmeans, Gaussian mixture model, EM algorithm,
- Ex MNIST data and comparison with CNN
- hierarchical patterns
Ideally, participating students should have successfully completed an introductory course in Artificial Intelligence. However, the course is self-contained and any student with a solid basis in Computer Science -- algorithms, data structures, programming, propositional logic, NP-hardness -- should in principle be able to follow. Prior knowledge about search (the A* algorithm etc) is an advantage and knowledge on Java programming.
probabilities and statistics, main distributions, probability density functions, matrix calculation, basic optimization (Newton's methods, Gradient methods...)
The math reminders will be done in general (the prerequisites having been seen in 1A and 2A), but very quickly.
AI prt (50%)
ML part (50%): 1h short written exam (25%), ML mini-project (25%)
The exam is given in english only
The course exists in the following branches:
Course ID : WEUMAIA0
You can find this course among all other courses.
• M. Ghallab, D. Nau and P. Traverso, “Automated Planning”, Morgan-Kaufman, 2004.
• S. Russell and P. Norvig, "Artificial Intelligence: A Modern Approach", chapter XI“, Prentice Hall, 2002
Trevor Hastie, Robert Tibshirani et Jerome Friedman (2009), "The Elements of Statistical Learning," (2nd Edition) Springer Series in Statistics
Christopher M. Bishop (2006), "Pattern Recognition and Machine Learning," Springer
Kevin P. Murphy (2012), "Machine Learning: a Probabilistic Perspective", The MIT Press
Date of update September 3, 2020