ADVANCED MACHINE LEARNING

Along the course a number of assignments will be proposed to be resolved individually. We only allow ”Type 1 collaboration”. This means that collaboration is allowed, but the final product must be individual. You are allowed to discuss the assignment with other team members and work through the problems together. What you turn in, however, must be your own product, written in your own handwriting, or in a computer file of which you are the sole author. Copying another’s work or electronic file is not acceptable.

– Assignment must be delivered on the established date. No assignment will be considered after deadline.
(A project work will be proposed to be resolved in groups of 2/3 students. The project will be evaluating according to the following criteria:

– A structured report (problem statement, background on models and techniques involved, data preparation, computational experiments, analysis of results).

Brief oral presentation of the project and discussion on the assignments.

2. Project [50%] +Oral exam [50%]

A project work will be proposed to be resolved in groups of 2/3 students. The project will be evaluating

according to the following criteria:

– A structured report (problem statement, background on decision models and techniques involved, data preparation, computational experiments, analysis of results).

Brief oral presentation of the project. An oral examination that will evaluate: Knowledge of Fundamental Concepts, Overall Understanding, Knowledge of specific models and methods, argumentation ability.

Valutazione:

Voto Finale

Obiettivi formativi

This machine learning advanced course is aimed especially for students who are already familiar with the basics of machine learning and wish to strengthen their knowledge and explore important advanced topics in order to posses in-depth and wide range capabilities at this so important field.

The course will cover some of the most important advanced topics in machine learning such deep learning and reinforcement learning, with their underlying theory but also a focus on modeling and practical implementation.

These advanced techniques will be applied to a number of applications, including: image recognition, natural language processing, recommendation systems.

Contenuti

Introduction to Deep Learning

Optimization techniques for training deep models

Convolutional Neural Networks

Unsupervised representation learning

Deep Learning for data sequences

Reinforcement learning

Programma esteso

Training Deep Networks:

Objective functions

Activation Functions

Regularization

Gradient-based optimization

Focus on Deep Networks:

Autoencoders

Convolutional Neural Networks

Recurrent and Recursive Networks

Practical Methodology:

Performance Metrics and baseline models

Selecting hyper-parameters

Reinforcement Laerning

Bibliografia consigliata

Ian Goodfellow and Yoshua Bengio and Aaron Courville, Deep Learning, MIT Press, 2016. http://www.deeplearningbook.org

Rasmussen, Gaussian Processes for Machine Learning, the MIT press 2006.

Further resource material will be made available on the e-learning platform.

Modalità di erogazione

Convenzionale

Metodi didattici

The course includes a part of theoretical lessons that will be held in the classroom and a part of exercises that will be held in the laboratory and / or classroom and which will require the use of your PC (or the one available at the University's computer labs) ).

The practical implementation of case studies will require the basic knowledge of R and Python programming languages.

Anno di corso: 1

Anno di corso: 2