Machine Learning

With the help of machine learning and its methods, you can train your neural networks. Marketing, FinTech, crisis management, risk management, security, medicine, robotics, etc. Areas of machine learning applications such as These are very popular and there is a great demand for each of them. The widespread stereotype that machine learning is the most difficult area of data analysis is just a myth. Our course will help you master the basics of artificial intelligence, update the mathematical framework in your memory, and provide a complete set of tools for working with neural networks.

Application to the program is currently not active
Start date



2.5 months/120 hours

Group size

25-30 students


I, III 18:30-21:30

Admission requirements

Readiness for intensive training

Language skills

Knowledge of English at least Intermediate level


To have a personal computer or a laptop


18 years and older

Upon the course completion you will::

Know and be able to use the basic machine learning algorithms (regression, decision tree, boost)

Be able to use algorithms in accordance with the task and the model

Analyze the project and analyze it step by step

Perform data preprocessing and data validation

Know the performance indicators of the model: Accuracy, Precision, Recall, etc.

Understand and practice unsupervised learning and deep learning with frameworks.


Machine Learning

Number of modules

 Probability (Data Distributions review), Bayesian statistics

  •        Matrix calculus(Take derivative from matrix)
  •        Matrix decomposotion techniques(SVD, LR)
  •        Math operations with numpy, efficient matrix operations, broadcasting, vectorization
  •        KDE
  •        Q-Q Plots

General Stages in ML project


  • Problem Definition
  • Research
  • Data Aggregation/Mining/Scraping
  • Data Preparation/Preprocessing
  • Modelling
  • Evaluation
  • Deployment
  • Data Preprocessing


Data Import (pandas, numpy)

  • Handling missing values
  • Categorical data encoding
  • Outlier detection
  • Date data processing (time series)
  • Dataset splitting
  • Cross Validation (Overfitting, Underfitting)
  • Feature Scaling: standardization, normalization
  • PCA + Theory

Confusion Matrix and Model Performance metrics


  • Recall
  • Precision
  • Accuracy
  • Root Mean Square Error
  • F1 score
  • Confusion Matrix
  • Gini Coefficient

Regression Algorithms (with Matrix notation):

  • Linear regression + Theory
  • Logistic regression + Theory
  • Regularization techniques (Lasso, Ridge)
  • Knn

Decision Tree Algorithms:

  • CART + Theory
  • Bagging & Boosting + theory
  • Random Forest + theory

Boosting Algorithms:

  • AdaBoost
  • XGBoost + Theory
  • LightGBM
  • CatBoost

Neural Networks

  • Neural Networks, backpropagation and forward propagation (single layer perceptron)
  • Multilayer perceptron
  • Cost functions
  • Activation functions

Optimization Methods (Hyper-parameters, Gradient Descent etc.)

  • Gradient Descent
  • Stochastic Gradient Descent

Hyperparameter optimization

  • Grid search
  • Random search
  • Bayesian optimization

     Imbalanced classification

  • ROC curve and threshold choice
  • SMOTE (Oversampling)
  • Near-Miss (Undersampling)

Model explainability

  • Permutation importance
  • Partial dependency plots
  • SHAP values

Unsupervised learning:

  • K_Means clustering + Theory
  • Clustering evaluation metrics
  • Dimensionality reduction (e.g. t-SNE, UMAP)

DL Frameworks (Tensorflow2 , Keras)

Introduction to Tensors, variables

Introduction to Gradients and Automatic Differentiation (tf.GradientTape())

Introduction graphs and functions

Introduction to modules, layers, and models

Basic training loops

Advanced Automatic Differentiation

Keras Sequential model

Training & evaluation with the built-in methods

Save and load Keras models

Working with preprocessing layers

Working with the model’s layers

Dropout, Weight initialization

Optimization algorithms (Mini batch gradient descent, Momentum, RMSProp, Adam)

Batch normalization

CNN (Convolutional Neural Network)



Fully connected

RNN (Recurrent Neural Networks)





Classical methods overview(language model, tf-idf vectorization)

Word embeddings, CBOW and skip-gram methods, word2vec