MVA-2021

My solutions of the homeworks of several courses of the Masters MVA (Mathématiques, Vision, Apprentissage)

First Semester

3D Computer Vision

Instructors: Pascal Monasse and Mathieu Aubry.

Includes:

• Panorama with homography computation and perspective projection
• Fundamental matrix computation with SIFT keypoints
• Disparity map computation by propagation of seeds
• Disparity map computation by Graph-Cuts

Object Recognition

Instructors: Jean Ponce, Ivan Laptev, Cordelia Schmid and Josef Sivic

Includes:

• Instance Level recognition
• Neural Networks
• Bird Classification Data-Challenge: Code & Report

Computational Statistics

Instructor: Stéphanie Allassonnière

Includes:

• Simple implementation of SGD
• Implementation of Gaussian-Mixtures Model with EM algorithm and Adaptative Importance Sampling with Monte-Carlo method
• Implementation of Hastings-Metropolis Sampler and some extensions (Adaptative HM sampler within Gibbs and Parallel Tempering)
• Implementation of Tempering SAEM from A New Class of EM Algorithms. Escaping Local Minima and Handling Intractable Sampling

Convex Optimization

Instructor: Alexandre d'Aspremont

Includes:

• Convexity, Conjugate Function
• Duality
• LASSO Regularization
• Final Exam

Reinforcement Learning

Instructors: Alessandro Lazaric and Matteo Pirotta

Includes:

• Dynamic Programming (MDP, Policy Iteration and Value Iteration)
• Approximate RL (TRPO, TD, REINFORCE, DQN)
• Exploration (UCB, Best Arm Identification, Bernoulli Bandits)
• Model Selection in Contextual Bandits

Second Semester

Point Clouds & 3D Modelling (NPM3D)

Instructors: François Goulette, Jean-Emmanuel Deschaud and Tamy Boubekeur

Includes:

• Basic Transformations of 3D Point Clouds
• Iterative Closest Point
• 3D Descriptors
• Surface Reconstruction
• Modeling and Shape Detection
• Deep Learning for Point CLouds

Deep Learning in Practice

Instructor: Guillaume Charpiat

Includes:

• Hyperparameter and training basics
• Visualization with grad-CAM
• Graph Neural Network
• Transfer Learning
• Learning Dynamical Systems
• Generative Models

Graphs in Machine Learning

Instructors: Daniele Calandriello and Michal Valko

Includes:

• Spectral Clustering
• SSL
• Spectral GNN

Kernel Methods

Instructors: Julien Mairal and Jean-Philippe Vert

Includes:

• Positive definite kernels and boundedness
• RKHS
• COCO: Covariance Kernel
• Bn-Splines and diffusion kernel
• Final exam (Positive Definiteness, Equivalence classes, Kernel mean embedding, Dot-product kernel)

Sequential Learning

Instructors: Pierre Gaillard and Rémi Degenne

Includes:

• Rock Paper Scissors with bandits and Bernoulli bandits