Seminar Course on Dynamic Decision Making
Code | Completion | Credits | Range |
---|---|---|---|
01DROS | Z | 2 | 0+2 |
- Garant předmětu:
- Lecturer:
- Tutor:
- Supervisor:
- Department of Mathematics
- Synopsis:
-
The seminar is devoted to the actual topics and trends in decision making, machine learning (ML) and artificial intelligence (AI). It will extend the topics learned in the lecture course 01DRO1, in particular formalisation of DM problem and its solution incl. techniques to tackle the problem; multi-agent DM and related tasks incl. possible ways of agents? interaction.
A sub-selection of relevant articles presented at the main DM, ML and AI conferences will be discussed.
- Requirements:
-
The seminar is intended for all students who attend the lecture course 01DRO1 and are interested in theoretic solutions and/or their subsequent practical use.
Individual work of students will consist of active participation to seminars, which is necessary condition for getting the allocated credits.
- Syllabus of lectures:
-
A part of the seminal will be devoted to practical tasks, in particular:
- Practical introduction into ML and business intelligence will be demonstrated on a real project. The seminar will connect various courses studied within MI or AMSM with practical applications (taxis, insurance, e-commerce,...), visualization in Qlik (Tableau), programming in R/Python.
- Description of workflow needed for using linear models and logistic regression (data preparation, pre-processing, scaling, etc.) will be illustrated on unfinished calls in an insurance company call centre.
- Practical use of decision trees, random forests, and gradient boosting technique will be demonstrated. Preparing a customer?s action for an insurance company will serve as a practical example.
- Processing of natural language will be demonstrated on searching potential employees using methods TFIDF and Word2vec serving for evaluation texts in CV databases.
- Introduction into futures trading incl. main principles, common strategies used and open problems.
- Concept of lazy learning for DM will be illustrated on a nontrivial problem of helicopter flight stabilisation.
- Syllabus of tutorials:
- Study Objective:
-
Acquired knowledge:
The elective seminar course focuses on using methods and algorithms of dynamical decision making under uncertainty (DM). Theoretical knowledge of DM will be complemented by examples of their practical use. This will enhance the students? ability to solve practical decision tasks and inspire further theoretic and algorithmic research.
Acquired skills:
The seminar will support understanding how to design elements and methods inevitable for optimised decision making.
- Study materials:
-
Compulsory literature:
[1] M. L. Puterman: Markov Decision Processes, Wiley, 1994. (selected chapters).
[2] R. S. Sutton, A. G. Barto: Reinforcement Learning: An Introduction MIT Press, Cambridge, 1998 (selected chapters).
Optional literature:
[3] S. French. Decision Theory. Halsted Press, 1986.
[4] L. Savage. The Foundations of Statistics. Wiley, 1954.
[5] D. P. Bertsekas. Dynamic Programming & Optimal Control, 1,2. Athena Scientific Press, 2005
- Note:
- Further information:
- No time-table has been prepared for this course
- The course is a part of the following study plans:
-
- Aplikované matematicko-stochastické metody (elective course)