Machine-Learning Techniques

---

Course structure

WEEK 2 - Models of regression; Linear regression - least squares; Polynomial regression - learning curves; Regularized linear models - Ridge, LASSO

WEEK 3 - Models of regression; Linear regression - least squares; Polynomial regression - learning curves; Regularized linear models - Ridge, LASSO

WEEK 4 - Models of classification ; Discriminant functions and decision boundaries - two classes, multiple classes, least squares, perceptron; Probabilistic generative and discriminative models - ML, Naive Bayes, exponential family, logistic regression

WEEK 5 - Models of classification ; Discriminant functions and decision boundaries - two classes, multiple classes, least squares, perceptron; Probabilistic generative and discriminative models - ML, Naive Bayes, exponential family, logistic regression

WEEK 6 - Models of classification ; Discriminant functions and decision boundaries - two classes, multiple classes, least squares, perceptron; Probabilistic generative and discriminative models - ML, Naive Bayes, exponential family, logistic regression

WEEK 7 - Models of classification ; Softmax regression ; k-NN - regression and classification problems

WEEK 8 - Support Vector Machines; Linear SVM - soft margin classification; Nonlinear SVM - kernels

WEEK 9 - Decision Trees; Training decision trees, making predictions;

WEEK 10 - Ensemble Methods and Random Forests; Bagging, Boosting;

WEEK 11 - Clustering; KMeans - algorithm, demo and how to select k-HAC

WEEK 12 - Neural networks; Multi-layer perceptron, activation functions; Training - SGD and back propagation; Hyperparameters - number of layers, neurons, activation functions.

---

What you’ll learn

• Demonstrating in-depth understanding of machine learning algorithms - model, objective or loss function, optimization algorithm and evaluation criteria.

• Tweaking machine learning algorithms based on the outcome of experiments - what steps to take in case of underfitting and overfitting.

• Being able to choose among multiple algorithms for a given task.

• Developing an understanding of unsupervised learning techniques.

---