Master 2020/2021

Machine Learning

Category 'Best Course for Career Development'

Category 'Best Course for Broadening Horizons and Diversity of Knowledge and Skills'

Type: Compulsory course (Machine Learning and Data Analysis)

Area of studies: Applied Mathematics and Informatics

Delivered by: Department of Informatics

Where: St. Petersburg School of Physics, Mathematics, and Computer Science

When: 1 year, 2-4 module

Mode of studies: offline

Open to: students of one campus

Instructors: Степанов Денис Вячеславович, Aleksei Shpilman

Master’s programme: Machine Learning and Data Analysis

Language: English

ECTS credits: 8

Contact hours: 112

Course Syllabus

Abstract

It is a compulsory discipline. The purpose of mastering the discipline "Machine Learning" is to develop students 'theoretical knowledge and practical skills on the basics of machine learning, mastering students' tools, models and methods of machine learning, as well as acquiring the skills of a data scientist and developer of mathematical models, methods and analysis algorithms data. As a result of mastering the discipline, the student must: - Know the key concepts, goals and objectives of using machine learning; methodological foundations of the application of machine learning algorithms. - Be able to visualize the results of machine learning algorithms, choose a machine learning method appropriate to the research task, and interpret the results. - Have the skills (gain experience) of reading and analyzing academic literature on the application of machine learning methods, building and evaluating the quality of models.

Learning Objectives

• 1. The formation of students' theoretical knowledge and practical skills on the basics of machine learning.
• 2. Students mastering tools, models and methods of machine learning
• 3. Acquiring the skills of a data scientist and developer of mathematical models, methods and algorithms for data analysis.

Expected Learning Outcomes

• Knows: basic concepts and tasks of machine learning and data analysis; basic principles, tasks and approaches, use in various fields of science and industry.
• Knows: general view of the metric classifier; K nearest neighbors algorithm; sampling algorithms.
• Has skills in algorithms. Knows: clustering algorithms with a fixed number of clusters; density clustering algorithms.
• Knows: rules and quality analysis (accuracy, completeness). Possesses analysis skills using the ROC curve. He knows the algorithm for constructing decision trees; informational gain criterion and Gini criterion.
• Owns the concepts of: perceptron and dividing hyperplane. Owns concepts: transition to space of increased dimension. Knows the support vector method
• Owns the concepts of: logistic regression; gradient descent; neural networks and gradient backpropagation algorithm
• Fluent in concepts: linear regression; polynomial regression; displacement and dispersion
• Has skills in algorithms. Owns the concepts of: Monte Carlo Searches; simulated annealing algorithm; genetic algorithm.
• owns concepts: voting; bootstrapping; boosting, adaptive boosting, gradient boosting.
• Owns concepts: ridge regression.

Course Contents

• Types of Machine Learning Tasks
  The subject and tasks of machine learning and data analysis. Basic principles, tasks and approaches, use in various fields of science and industry. The main stages of the evolution of machine learning algorithms.
• Clustering Algorithms
  Clustering algorithms with a fixed number of clusters.
• Linear Classifiers
  Perceptron and dividing hyperplane. Transition to space of increased dimension. Support Vector Method
• Clustering Algorithms
  Density clustering algorithms. Hierarchical clustering.
• Metric classifiers
  General view of the metric classifier. Algorithm K nearest neighbors. Pattern selection algorithms.
• Decision trees
  Rules and quality analysis (accuracy, completeness). Analysis using the ROC curve. Algorithm for constructing decision trees. Informational gain criterion and Gini criterion. Forests of decisive trees.
• Neural networks and deep learning
  Logistic regression. Gradient descent. Neural networks and gradient backpropagation algorithm. Deep learning, convolution and pooling.
• Regression analysis
  Ridge regression.
• Regression analysis
  Linear regression. Polynomial regression. Displacement and dispersion.
• Ensemble Methods
  Voting. Bootstrapping. Boosting, adaptive boosting, gradient boosting.
• Stochastic search
  Monte Carlo search. Simulated Annealing Algorithm. Genetic algorithm.

Assessment Elements

• Homework №1
• Homework №2
• Homework №3
• Exam (3 module)
  Экзамен проводится офлайн.
• Exam (4 module)

Interim Assessment

• Interim assessment (3 module)
  0.41 * Exam (3 module) + 0.59 * Homework №1
• Interim assessment (4 module)
  0.5 * Exam (4 module) + 0.25 * Homework №2 + 0.25 * Homework №3

Bibliography

Recommended Core Bibliography

• Флах П. - Машинное обучение. Наука и искусство построения алгоритмов, которые извлекают знания из данных - Издательство "ДМК Пресс" - 2015 - 400с. - ISBN: 978-5-97060-273-7 - Текст электронный // ЭБС ЛАНЬ - URL: https://e.lanbook.com/book/69955

Recommended Additional Bibliography

• James, G. et al. An introduction to statistical learning. – Springer, 2013. – 426 pp.