Machine Learning Basics (MSc)
Lecture Slides
Lecture Notes
AI-900 materials (for AI-900
Microsoft Azure AI Fundamentals Certification)
Exercises
Topics:
1. Basic concepts
2. Linear Algebra
3. Probability and Information Theory
4. Numeric Computations
5. Data Preprocessing
6. Dimensionality Reduction
7. Regression Models 1
8. Regression Models 2
9. Classification 1
10. Classification 2
11. Clustering 1
12. Clustering 2
13. Association Rule Learning
14. Reinforcement Learning
Literature:
ˇ W. McKinney: Python for Data Analysis (1 ed.). O'Reilly Media, Inc. 2012.
ˇ Christopher Bishop: Pattern Recognition and Machine Learning, Springer, 2006.
ˇ D. Conway, J.M. White: Machine Learning for Hackers, O'Reilly Media, Inc., 2012.
ˇ I. Goodfellow, Y. Bengio, A. Courville: Deep Learning, MIT Press, 2016.
Main concepts:
Supervised Learning, Unsupervised Learning, Linear Regression (one/multiple variables), Gradient Descent, Feature Normalization, Polynomial Regression, Normal Equation, Logistic Regression, Binary Classification, Multiclass Classification (One-vs-all), Regularization (overfitting, underfitting), Regularized Linear Regression, Regularized Logistic Regression, Neural Networks, Backpropagation Algorithm, Gradient Checking (numerical), Train/Validation/Test Sets, Diagnosing Bias vs. Variance, Regularization and Bias/Variance, Learning Curves (training set size), Error Metrics for Skewed Classes, Support Vector Machine, Kernels in SVM, Clustering, K-Means Algorithm, Choosing the Number of Clusters, Dimensionality Reduction (PCA), Anomaly Detection, Gaussian Distribution, Multivariate Gaussian Distribution, Recommender Systems, Content Based Recommendations, Collaborative Filtering, Stochastic Gradient Descent, Mini-Batch Gradient Descent, Map Reduce and Data Parallelism
Programming tasks:
- SPAM filtering (larger database)
- Anomaly detection vs supervised learning
- Transforming data to Gaussian with random sample
- K-means clustering with adding new clusters
- Neural nets for logical operators
- Handwritten characters classification
- Recommendation system for own dataset