Machine Learning #1 and #2

This repo contains materials for the introductory/intermediate Machine Learning (ML) courses (4 credits total) taught in the MSc in Business Analytics program at the Central European University (CEU) starting January 2018. The material for the previous 2-credit course taught in 2016 and 2017 can be found here.

Course Description and Objectives

The breakdown below (into ML #1 an #2) is somewhat adhoc and mainly to comply with administrative requirements. Both courses will intertwine general ML concepts, algorithms and software implementations/tools and will aim to strike a balance of theory and practice with the goal of equiping students with both the foundations to understand the ML methodology and also with the skills needed for using ML in practical business applications.

Data Science and Machine Learning 1 (Concepts):

After an overview of the entire data science landscape this course will focus on machine learning. The course will introduce the main fundamental concepts in machine learning (supervised learning, training, scoring, accuracy measures, test set, overfitting, cross validation, model capacity, hyperparameter tuning, grid and random search, regularization, ensembles, model selection etc.) The concepts will be illustrated with R code therefore it requires prior familiarity with R.

Data Science and Machine Learning 2 (Tools):

This course will build on the previous one (which introduced the basic concepts in machine learning) and will discuss state-of-the-art algorithms for supervised learning (linear models, lasso, decision trees, random forests, gradient boosting machines, neural networks, support vector machine, deep learning etc.). A large part of the course will be dedicated to using (hands-on) the software tools for machine learning used by data scientists in practice (various high-performance R packages, h2o, xgboost, libraries for deep learning on GPUs etc.).

Course Designer (and Instructor previous years):

Szilárd Pafka

Instructors

Zoltán Papp
Jenő Pál (TA)

Grading

• 50% Weekly Assignments (Homework Exercises)
• 50% Final Exam

(each course ML #1 and #2 separately)

Announcements and Q&A

Class announcements and student Q&A will be done via github issues.

Syllabus and Schedule:

ML #1

Week 1: Overview of data science. The elements of a data science project. Exploratory data analysis. Data preparation/munging. Data visualization. Machine learning. Workflow, reproducibility and productivity. Lecture | Lab.

Week 2: Introduction to supervised learning. Linear models vs k-nearest neighbors. Training and test error. Bias and variance. Lecture | Lab.

Week 3: Model evaluation and selection. Overfitting, regularization, cross-validation. ROC curve, AUC. Hyperparameter tuning, grid and random search. Lecture | Lab.

Week 4: Unsupervised learning. Clustering (k-means, hierarchical). Final Exam (ML #1). Lecture | Lab.

ML #2

Week 9: Linear models, lasso. Trees, random forests and gradient boosting machines. Tools: R packages, Vowpal Wabbit, xgboost, lightgbm, H2O Lecture | Lab.

Week 10: Support vector machines. Neural networks and deep learning. Tools: R packages, H2O, Keras. Lecture | Lab.

Week 11: Ensembles, stacking. Deploying machine learning models to production. Lecture | Lab.

Week 12: Recap and summary. Final Exam (ML #2). Lecture | Lab.