Metody Inteligencji Sztucznej i Obliczeniowej

This repository contains resources necessary to complete Artificial Intelligence courses (MISiO) at Poznan University of Technology.

• Grades 2020
• Grades 2019

Python Package

To run some of the tasks a special python package will be needed. To install it use:

git clone https://github.com/mihahauke/misio_labs
cd misio
sudo pip3 install .

or:

sudo pip3 install git+https://github.com/mihahauke/misio_labs

or if you don't have root access:

pip3 install git+https://github.com/mihahauke/misio_labs --user 

Grading

tl;dr:

• you have to do all tasks
• delays are penalized
• cheating and sharing code will result in grade 2/2

Full

"Unless stated otherwise in the task description" applies to all of the following points.

• each task has to be completed to pass the course; failure to do so will result in FAILING (getting grade 2) regardless of points scored on other tasks;
• a task is considered completed if a solution is submitted and it scores at least 30% of achievable points (before potential delay penalty);
• attendance is officially mandatory (as per University code); in case of any doubts about scoring/cheating lack of attendance may result in disadvantageous consideration
• solutions submitted before the start of the next class; every started week of delay results in -20% penalty (sometimes you'll have 2 weeks)
• sharing your code or solutions is prohibited (you may however share your thoughts and ideas)
• submitting someone else's solutions or its parts will result in grade 2/2 and any legal repercussions available

In case of FAILING:

There are 2 options:

• 2/3 - complete every task (except for AC) for at least 30% and get >= 50% points in total (no penalties for delays)
• 2/X - complete every task (including AC) and you get whatever is determined by your points (no penalties for delays)

Lab 1 Intelligent agents (10 pts)

Introduction to artificial intelligence basics: agents, environments, rationality etc. This task uses python's aima3 library for AIMA(Artificial Intelligence Modern Approach).

Lab 2 Uncertain Wumpus (15pts)

Practical example of uncertainty in Artificial Intelligence.

Lab 3 Lost Wumpus (15pts)

Using histogram filter for navigation.

Lab 4 MDP AIMA (10pts)

Introduction to Markov Decision Processes (MDPs) which are a fundamental framework for modern AI algorithms.

Lab 5 MDP Wumpus Store (15pts)

More Markov Decision Processes.

Lab 6 Q-Learning(15pts)

Reinforcement Learning using one of the most popular and well known algorithm: Q-learning

Lab 7 Actor-critic(15pts)

Reinforcement Learning: actor-critic (continuous action-spaces)

Authors

• Michał Kempka

Acknowledgements

Big portions of the cirriculum was designed by Wojciech Jaśkowski and based on AI course on Berkley.