/ml

Primary LanguageJupyter Notebook

Training materials for the course "Machine learning"

Russia, Ulyanovsk Ulyanovsk State Technical University

1. Basics of Machine Learning

  • Brief history of basic technologies
  • Definitions
  • Basics of ML

2. Python

  • Basic syntax
  • Arithmetic
  • Strings
  • Lists
  • Exploratory data analysis with Pandas
  • Data visualization
  • Kaggle competitions
  • Titanic task

3. Linear regression

  • Linear regression
  • Cost function
  • Gradient descent
  • Linear regression with multiple variables
  • Debug (learning rate)
  • Normal equation
  • Features and polynomial regression
  • Multi-class classification
  • Feature scaling

4. Decision trees. KNN. Logistic regression. Regularization.

  • Decision tree
  • K-nearest neighbors
  • Classification problem
  • Sigmoid function
  • Decision boundary
  • Cost function
  • Regularization. Problem of overfitting

5. Machine learning system design

  • Error analysis. Metrics
  • Evaluating hypothesis. Train / test / validation set
  • High bias / high variance (model selection, regularization, learning curves)
  • Feature extraction:
  • Images
  • GEO
  • Date and time
  • Timeseries
  • Texts. One-hot encoding

6. Clustering

  • Clustering (k-means, c-means, hierarchical clustering)
  • Principal component analysis

7. Naive bayes and SVM

  • Bayes theorem
  • Naive bayes
  • Support vector machines

8. Neural networks

  • Non-linear hypothesis
  • Neurons and the brain
  • Forward propagation (XNOR example)
  • Back propagation
  • Parameters initializing

9. Deep learning. Convolutional neural networks

  • Convolution. Feature representation as hierarchy
  • Filters, stride, padding
  • Pooling
  • Popular architectures: AlexNet, VGG, ResNet,
  • Classification, localization, regression

10. Recurrent neural networks

  • Basics of recurrent NN
  • LSTM
  • Time-series analysis
  • Text analysis

11. Time seriaes analysis

12. Trees

  • Decision trees
  • Random forest
  • XGBoost
  • CatBoost

Links

Visual Attention Model