arturbalcz / misio-labs

Solution to tasks from Artificial Intelligence courses (MISiO) at Poznan University of Technology.

Metody Inteligencji Sztucznej i Obliczeniowej

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

• Lectures
• [Grades 2021](Coming soon)
• 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

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 cummulative -20% penalty, some examples:
  • 15/15 (100%), 5 weeks of delay, task accepted, points gained: 15 (1-0.25) = 0
  • 10/15 (66.66%), 2 weeks of delay, task accepted, points gained: 10 (1-0.22) = 6
  • 4.5/15 (30%), 1 week of delay, task accepted, points gained: (1-0.2*1) *4.5 = 3.6
  • 4.5/15 (30%), 2 weeks of delay, task accepted, points gained: (1-0.2*2)*4.5 =2.7
  • 4/15 (<30%), no delay, task not accepted, Failing grade (2),
• 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:

Complete every task 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 in continuous spaces (can be solved with actor-critic algorithm or e.g. ddpg)

Authors

• Michał Kempka

Acknowledgements

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