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.
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.
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.
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, xgboost, libraries for deep learning etc.).
Szilárd Pafka
Zoltán Papp
Jenő Pál (TA)
- 45% Weekly Assignments (homework exercises). These will be submitted using Moodle.
- 45% Final Exam
- 10% Quizzes at the beginning of each lecture, except the first lectures of each course. Missing a lecture or being late will result in 0% for the actual quiz score.
(each course ML #1 and #2 separately)
Weekly assignment dates and deadlines for ML #1:
- Jan 16, due on Jan 29 8:00 AM
- Jan 23, due on Feb 5 8:00 AM
- Jan 30, due on Feb 12 8:00 AM
Weekly assignment acceptance policy and achievable grades:
- 100% until due date
- 50% within 24 hours past due date
- 0% after that.
Class announcements and student Q&A will be done via Moodle
Week 1: Overview of data science. The elements of a data science project. Data preparation. Exploratory data analysis, 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. Validation set. Bias and variance. Model evaluation and selection. Benchmark model selection. Boostraping, resampling. Time series split up by time. Lecture | Lab.
Week 3: A high-level introduction to RF, Lasso, SVM, Boost, NN. Overfitting, regularization, cross-validation, naive implementation. ROC curve, AUC. Calibration plot. Hyperparameter tuning, grid, and random search. Lecture | Lab.
Week 4: Unsupervised learning. Clustering (k-means, hierarchical). PCA / ICA. Time Series Analysis: correlation, autocorrelation, Filtering. Reviewing best practices in data science and practical research. Lecture | Lab.
Week 10: Deeper understanding of linear models, lasso. Understanding and tuning parameters for trees, random forests and gradient boosting machines. Impact of correlated features. Support vector machines. Tools: R packages, xgboost, lightgbm Lecture | Lab.
Week 11: Neural networks and deep learning. Reinforcement Learning. Evolutionary Computing. Tools: R packages, Keras. Lecture | Lab.
Week 12: Ensembles, Stacking. Deploying machine learning models to production. Lecture | Lab.
Week 13: Recap and summary. Final Exam (ML #1 and #2). Lecture | Lab.