Projects

In this project, you will implement the learning of concepts expressible by s-CNF formulas. You should use the basis expansion trick in order to reduce this problem to the learning of monotone conjunctions. Note, that the environment in this assignment does not provide samples independently and the observations can be incomplete in the sense discussed on lectures. Your agent should work within the mistake-bound setting also discussed on the lectures.

The instructions and code are included in project1_2021.zip. This file is meant to be self-contained and includes all the detailed instructions. Submit your solutions (the .ipynb file with your modifications) to the BRUTE upload system. If you have any questions, contact Jáchym Barvínek on MS Teams or via e-mail barvijac@fel.cvut.cz.

The project is due in three weeks after it's assigned. The maximum amount of points you can get from this project is 12.5 (Which you can see as 25 with weight 0.5 in BRUTE).

In this project, your task is to i) complete FOL based agent who uses clauses and background knowledge, ii) present one learning scenario to demonstrate the agent's learning ability. Clone the project https://gitlab.fel.cvut.cz/svatoma1/assignment-2nd-smu-2021, where you will find all instructions. The project is due in three weeks after it's assigned (see the deadline in BRUTE). The maximum amount of points you can get from this project is 12.5 (which you can see as 25 with a 0.5 weight in BRUTE).

In case of any question, contact Martin Svatos by e-mail* or use MS Teams.

* use university e-mail; other addresses may end up in spam

In this project, we will try to derive an optimal strategy for a simplified version of the blackjack game using the SARSA algorithm. Follow detailed instructions in this file. The base for your code can be found in this archive.

If you decide to submit late, you are likely to submit during the examination period. I cannot guarantee that your assignment will be graded if you submit it less than two working days before the intended exam date.

This project is scored only for students who opted out of the 3-project grading. However, everyone is welcome to try solve it to cement their understanding of factor-based MAP inference in Bayes Networks. All instructions are in the code that is to be unzipped and uploaded to Colab. The solutions are due in 3 weeks, i.e., on May 28.