Program Overview

Introduction to Machine Learning

Course Annotation

This course covers the basics of data analysis using machine learning methods. It explores linear models, decision trees, and neural networks, with a primary focus on supervised learning, as well as algorithms for unsupervised and reinforcement learning. The working language of the course is Python, making it beneficial for students planning to participate in data analysis of physical experiments.

Authors and/or Instructors

Evgeny Yurievich Soldatov

• Academic Degree: Candidate of Physical and Mathematical Sciences
• About the Author/Instructor: Associate Professor of the Department of Nuclear Physics and Technology (Department of Elementary Particle Physics No. 40). Deputy head of the NIYAU MIFI group in the ATLAS experiment (CERN), coordinator of physical data analysis.
• Scientific Interests: Experimental physics of elementary particles, electroweak theory, deviations from the Standard Model, cosmology, statistical data analysis, machine learning.

Alexey Viktorovich Grobov

• Academic Degree: Candidate of Physical and Mathematical Sciences
• About the Author/Instructor: Leading Researcher at the Kurchatov Institute. Associate Professor of the Department of Elementary Particle Physics No. 40. Participant in international experiments searching for dark matter, DarkSide, DEAP-3600.
• Scientific Interests: Physics of rare processes, machine learning and data analysis, low-background detectors, dark matter, nuclear physics, cosmology.

Course Topics

1. Introduction and Toolkit
   • Introduction to the course, its goals, and objectives. The subject of machine learning. Brief introduction to the necessary toolkit for classes: Python, Jupiter Notebook, etc.
2. Basic Knowledge of Mathematical Statistics
   • Knowledge from mathematical statistics and probability theory. Types of distributions of random variables. The most important theorems and methods of mathematical statistics. Concepts of hypothesis and criterion, examples of use. Bayesian approach.
3. Data Analysis
   • Data collection and processing. Correlations and dependencies in data. Pearson's criterion. Kolmogorov-Smirnov criterion. Genetic algorithm and gradient descent. Supervised learning. Classification and regression tasks.
4. Linear Algorithms
   • Method of least squares. Method of maximum likelihood. Error functional. Linear regression, logistic regression. Fisher discriminant. Quality metrics, ROC curve.
5. Decision Trees
   • Decision trees and random forests. Regularization and parameter tuning. Gradient boosting. Training stop criteria.
6. Cross-Validation
   • Overfitting. Regularization and cross-validation. Feature selection, scaling, dimensionality reduction. Principal component analysis.
7. Neural Networks
   • Algorithm for the operation of neural networks. Feedforward neural networks. Backpropagation error method. Universal approximation theorem. Multidimensional perceptron. Complex models, stacking.
8. Convolutional Neural Networks
   • Image processing. Convolutional neural networks. Calculations on GPU.
9. Unsupervised Learning
   • Clustering algorithms. Curse of dimensionality. Independent component analysis.
10. Reinforcement Learning
    • Robot training. Moving averages. Value optimization strategies.