Advanced Machine Learning

You are here

Home » Studies » Masters » Master in Data Science » Curriculum » Syllabus » Advanced Machine Learning
Credits
6
Types
Elective
Requirements
This subject has not requirements, but it has got previous capacities
Department
CS
The goal of machine learning is the development of theories, techniques and algorithms that allow a system to modify its behavior through observational data that represent incomplete information about a natural process or phenomenon, subject to
statistical uncertainty. Machine learning is a meeting point from different disciplines: multivariate statistics, algorithms and mathematical optimization, among others.

The subject reviews some foundations and then delves into various modern non-linear learning techniques ranging from modern neural networks to advanced kernel-based learning methods and the latest developments in ensemble methods. It also aims to provide a rather unified view of the area and possible future prospects.

Teachers

Person in charge

  • Luis Antonio Belanche Muñoz ( )

Others

  • Jamie Arjona Martínez ( )

Weekly hours

Theory
3.2
Problems
0
Laboratory
1
Guided learning
0
Autonomous learning
7.38

Competences

Transversal Competences

Information literacy

  • CT4 - Capacity for managing the acquisition, the structuring, analysis and visualization of data and information in the field of specialisation, and for critically assessing the results of this management.

Third language

  • CT5 - Achieving a level of spoken and written proficiency in a foreign language, preferably English, that meets the needs of the profession and the labour market.

Basic

  • CB6 - Ability to apply the acquired knowledge and capacity for solving problems in new or unknown environments within broader (or multidisciplinary) contexts related to their area of study.
  • CB7 - Ability to integrate knowledge and handle the complexity of making judgments based on information which, being incomplete or limited, includes considerations on social and ethical responsibilities linked to the application of their knowledge and judgments.
  • CB10 - Possess and understand knowledge that provides a basis or opportunity to be original in the development and/or application of ideas, often in a research context.

Generic Technical Competences

Generic

  • CG2 - Identify and apply methods of data analysis, knowledge extraction and visualization for data collected in disparate formats

Technical Competences

Especifics

  • CE3 - Apply data integration methods to solve data science problems in heterogeneous data environments
  • CE5 - Model, design, and implement complex data systems, including data visualization
  • CE8 - Extract information from structured and unstructured data by considering their multivariate nature.
  • CE9 - Apply appropriate methods for the analysis of non-traditional data formats, such as processes and graphs, within the scope of data science
  • CE10 - Identify machine learning and statistical modeling methods to use and apply them rigorously in order to solve a specific data science problem
  • CE11 - Analyze and extract knowledge from unstructured information using natural language processing techniques, text and image mining

Objectives

  1. Advanced machine learning methods
    Related competences: CT4, CT5, CE8, CE9, CE10, CB6, CB10,
  2. Bayesian statistics
    Related competences: CT4, CT5, CE5, CE8, CE10, CB7,
  3. Optimization of neural networks and support vector machines
    Related competences: CT4, CT5, CG2, CE3, CE5, CE11, CE8, CE9, CE10, CB6, CB7, CB10,
  4. Linear models and generalized nonparametric linear models for regression
    Related competences: CT5, CE5, CE10, CB10,
  5. Data cleaning
    Related competences: CT4, CG2, CE3, CE11, CE8, CE9, CB6,

Contents

  1. Theoretical refresher of machine learning. Introduction to Bayesian machine learning.
    Introduction to Bayesian thinking for machine learning. Learning by solving a regularized problem. Illustrative example.
  2. Learning in functional spaces
    Reproducing kernel Hilbert spaces. The representer theorem. Example 1: Kernel ridge regression. Example 2: The Perceptron and the kernel Perceptron.
  3. Fundamental kernel functions in R^d.
    Description and demonstration of fundamental kernel functions in R^d. Polynomial and Gaussian kernels. General properties of kernel functions.
  4. The support vector machine for classification, regression and novelty detection
    The support vector machine (SVM) is the flagship in kernel methods. Its versions for classification, regression and novelty detection are fully explained and demonstrated.
  5. Kernel functions for diferent data types
    Some kernel functions for different data types are presented and demonstrated, such as text, trees, graphs, categorical variables, and many others.
  6. Other kernel-based learning algorithms
    Additional kernel-based learning methods are explained, such as kernel PCA and kernel FDA. These are illustrated in several application examples.
  7. Introduction to deep neural networks. Autoencoders and Variational Autoencoders.
    Introduction to deep neural networks: reminder of fundamental neural network theory and optimization, qualitative description, loss functions, activation functions, regularization and best practices.
    Autoencoders and Variational Autoencoders.
  8. Special networks: (New) Hopfield neural networks and KANs
    Special networks: (New) Hopfield neural networks and KANs.
  9. Ensemble methods: baggers, boosters and stackers
    This activity cover the basic and modern developments in ensemble methods, including baggers, boosters and stackers.
  10. Advanced and hybrid techniques in deep networks and kernel methods
    Other methods are briefly introduced, such as the RVM and GPs. Nyström acceleration and Random Fourier features. Deep kernel learning and maybe others.

Activities

Activity Evaluation act




Final exam


Objectives: 1 2 4 3 5
Week: 17
Theory
3h
Problems
0h
Laboratory
0h
Guided learning
0h
Autonomous learning
18h


Teaching methodology

The course delves into the most important machine learning paradigms with a solid foundation in probability, statistics and math. The theory is introduced in lectures where the teacher exposes the concepts. These concepts are put into practice in the laboratory classes, in which the student learns to develop machine learning solutions to real problems of a certain complexity.

Students have to work and deliver a term project.

Evaluation methodology

The course is graded as follows:

F = Grade of the final exam
P1, P2, P3 = Grade of the practical works (1, 2 and 3)


Final grade = 25% F + 25% P1 + 25% P2 + 25% P3

Bibliography

Basic:

Previous capacities

Machine Learning course