Class Notes: Machine Learning
1. Introduction
Machine Learning (ML) is a subset of Artificial Intelligence that enables systems to learn from data
and improve performance without explicit programming. Goals include prediction, classification,
pattern recognition, and decision-making.
2. Types of Machine Learning
Supervised Learning: Input + labels → algorithms learn mapping. Examples: Linear Regression,
SVM.
Unsupervised Learning: No labels → discover patterns. Examples: K-means, PCA.
Reinforcement Learning: Agent learns via rewards and penalties. Examples: Q-learning.
3. Key ML Concepts
Dataset: training, validation, test sets.
Features (X): input variables. Labels (Y): outputs.
Overfitting: model learns noise. Underfitting: too simple model.
4. Common Algorithms
Regression: Linear, Logistic.
Classification: Decision Trees, Random Forest, SVM, Naïve Bayes, Neural Nets.
Clustering: K-means, Hierarchical, DBSCAN.
Dimensionality Reduction: PCA, t-SNE, LDA.
5. Model Evaluation Metrics
Classification: Accuracy, Precision, Recall, F1, ROC-AUC.
Regression: MSE, MAE, R².
Clustering: Silhouette Score, Davies-Bouldin Index.
6. Training & Optimization
Gradient Descent minimizes loss function.
Learning Rate controls step size.
Regularization: L1 (Lasso), L2 (Ridge).
Cross-validation improves generalization.
7. Neural Networks & Deep Learning
Perceptron: basic unit with activation function.
Activation functions: Sigmoid, ReLU, Tanh, Softmax.
Deep Learning: CNNs for images, RNNs/LSTMs for sequences.
8. Applications of ML
Image recognition, NLP, Fraud detection, Healthcare, Autonomous driving, Recommendation
systems.
9. Challenges in ML
Data quality, bias, interpretability, computation cost, privacy concerns.
10. Future Trends
Explainable AI, Federated Learning, Edge AI, AI + IoT/Robotics.
