Mathematics and Machine Learning for Data Science - course RUB 50,040. from SkillFactory, training 5.5 months, Date: August 13, 2023.

Mathematics course program

Part 1 - Linear algebra

• We study vectors and types of matrices
• Learning to perform operations on matrices
• Determining linear dependence using matrices
• We study inverse, singular and non-singular matrices
• We study systems of linear equations, eigen and complex numbers
instagram viewer
• Mastering matrix and singular decomposition
• Solving linear dependence problems using matrices
• Optimizing using the principal component method
• Reinforcing the mathematical foundations of linear regression

Part 2 - Basics of mathematical analysis

• We study functions of one and many variables and derivatives
• Mastering the concept of gradient and gradient descent
• Training in optimization problems
• We study the Lagrange multiplier method, Newton's method and simulated annealing
• We solve problems of predicting and searching for a winning strategy using derivative and numerical optimization methods
• Reinforcing the math behind gradient descent and simulated annealing

Part 3 - Fundamentals of probability and statistics

• We study the general concepts of descriptive and mathematical statistics
• Mastering combinatorics
• We study the main types of distributions and correlations
• Understanding Bayes' Theorem
• Learning a Naive Bayes Classifier
• We solve problems of combinatorics, validity and forecasting using statistics and probability theory
• We consolidate the mathematical foundations of classification and logistic regression

Part 4 - Time series and other mathematical methods

• Introducing Time Series Analysis
• Mastering more complex types of regressions
• Forecasting the budget using time series
• Reinforcing the mathematical foundations of classic machine learning models

Brief course program on Machine Learning

Tutor assistance during training

Module 1 - Introduction to Machine Learning

We get acquainted with the main tasks and methods of machine learning, study practical cases and apply the basic algorithm for working on an ML project
We solve 50+ problems to reinforce the topic

Module 2 - Data Preprocessing Methods

We study data types, learn to clean and enrich data, use visualization for preprocessing and master feature engineering
We solve 60+ problems to reinforce the topic

Module 3 - Regression

We master linear and logistic regression, study the limits of applicability, analytical inference and regularization. Training regression models
We solve 40+ problems to reinforce the topic

Module 4 - Clustering

We master learning without a teacher, practice its various methods, work with texts using ML
We solve 50+ problems to reinforce the topic

Module 5 - Tree-based algorithms: introduction to trees

Let's get acquainted with decision trees and their properties, master trees from the sklearn library and use trees to solve a regression problem
We solve 40+ problems to reinforce the topic

Module 6 - Tree-based algorithms: ensembles

We study the features of tree ensembles, practice boosting, use the ensemble to build logistic regression
We solve 40+ problems to reinforce the topic
We are participating in a competition on kaggle for training a tree-based model

Module 7 - Assessing the Quality of Algorithms

We study the principles of sample splitting, under- and overfitting, evaluate models using various quality metrics, learn to visualize the learning process
Evaluating the quality of several ML models
We solve 40+ problems to reinforce the topic

Module 8 - Time Series in Machine Learning

Let's get acquainted with time series analysis in ML, master linear models and XGBoost, study the principles of cross-validation and parameter selection
We solve 50+ problems to reinforce the topic

Module 9 - Recommender Systems

We study methods for constructing recommender systems, master the SVD algorithm, evaluate the quality of recommendations of the trained model
We solve 50+ problems to reinforce the topic

Module 10 - Final Hackathon

We apply all the studied methods to obtain maximum accuracy of model predictions on Kaggle