These Matlab laboratories were developed when I attended the course "055047 - ONLINE LEARNING AND MONITORING" during my studies at Politecnico di Milano
The original lab codes have been provided by the course professors Boracchi and Trovò, and originally had some missing parts to be completed. My task was to complete the algorithms code using what I learned from the lessons.
In particular these labs deals with the following topics of Online Machine Learning :
- Anomaly Detection and Domain Adaption
- Multi Armed Bandit algorithms and Expert Learning algorithms
- LAB 1/2/3/4 : set of techniques for learning in non-stationary environments. The goal is to detect a change of the input distribution in a stream of data and to consequently adapt the online learner to those changes, in order to improve the predition accuracy of the new incoming samples.
- LAB 5 : application of some online learning algorithms in the case we have a set of experts that gives feedbacks after the prediction (Expert Learning)
- LAB 6 : application of some online learning algorithms in the case we have a limited feedback (Multi Armed Bandit --> Thompson Sampling and UCB1 algorithms)
- LAB 7 : detection of anomalous heart signals, using PCA for dimensionality reduction of the initial dataset
- LAB 8 : application of some online learning algorithms in the case we have a limited feedback and a non-stationary environment (Non-Stationary Multi Armed Bandit)
- LAB 9 : an extension of LAB 8 where I implement some advanced NS-MAB algorithms, in order to show them during my oral examination