The objectives of the course include (a) the acquisition of specialized knowledge to address issues that arise in real-world applications (class imbalance, unequal classification error costs, limited training data, large data sets, data streams, data with multiple labels, instances and relations), (b) the acquisition of useful skills for researchers and practitioners (reading, writing and evaluating scientific publications, performing comparative analysis of machine learning algorithms, using data analysis software, implementing learning algorithms), and (c) acquaintance with modern popular applications (opinion mining, data mining for the retail business). The ultimate goal is to prepare students for both a professional career and the continuation of their studies at the third cycle.
Course Content (Syllabus)
Supervised ensemble methods, cost-sensitive learning, class imbalance, comparative evaluation of learning algorithms, multi-label learning, multi-instance learning, active learning, reading, evaluating and writing scientific publications, relational data mining, learning from data streams, opinion mining, data mining in the retail business.
Course Bibliography (Eudoxus)
Παρουσιάσεις μαθήματος, λίστα διαφορετικών συγγραμάτων ή/και επιστημονικών δημοσιεύσεων ανά αντικείμενο του μαθήματος. Ενδεικτικά:
- Zhou, Z.H. (2012) Ensemble Methods: Foundations and Algorithms, Boca Raton, FL: Chapman & Hall/CRC, 2012.
- Nathalie Japkowicz and Mohak Shah. 2011. Evaluating Learning Algorithms: A Classification Perspective. Cambridge University Press, New York, NY, USA.
- Burr Settles, Active Learning, Synthesis Lectures on Artificial Intelligence and Machine Learning
Morgan & Claypool Publishers, June 2012.