Course prepared for Sofia University: Palo Alto facility, 2024.
- (2024-03-31) Starting 2024 Spring Session
| N | Lecture | Desctription |
|---|---|---|
| 01 | Introduction | Introduction. Course logistics and syllabus. Historic reference. ML Paradigms |
| 02 | Supervised Learning | Supervised Learning Setting. Objects' features. Model outputs. Loss functions. Cross-validation. Hyperparameters tuning |
| 03 | ML Model Fitting Problems | Empirical and Structural Risk. Risk Minimization. Model Selection. Underfitting and overfitting. Error Decomposition: Bias-Variance Tradeoff and Double Descent |
| 04 | k-NN for Classification | Non-parametric Classification: k-NN Method and its variants (Euclidean and Manhattan Distance, 1-NN and k-NN, Weighted K-NN, Selection of Templates). Common Metrics. Classification Mean Error |
| 06 | k-NN for Regression | Non-parametric Regression: k-NN Method and its variants. Nadaraya-Watson kernel regression. Bias-Variance trade-off for k-NN Regression. Mean (Absolute) Test Error |
| 07 | Linear Regression | Linear Regression and its variants (Ridge, LASSO, Elastic Net). Polynomial Regression. Least Squares method. ML and MAP principles. Regression Quality Metrics |
| 08 | Classification Metrics | Binary Classification definitions. Accuracy. Confusion Matrix: TP, FP, TN, FN (and TPR, FPR, TNR, FNR). ROC / AUROC. Precision and Recall. PRC / AUPRC. Multi-class Classification variants and Class Imbalance |
| 09 | Exam | Final Exam: information and logistics. ML Pipeline Design topics. ML Concepts topics. ML Calculations |
| 10 | ML Buzzwords | AI/ML/DL "Buzzwords": broad concepts, research directions, state-of-the-art approaches |