The source code for the Bachelor degree in Computer Science.
We apply Deep Reinforcement Learning techniques to solve popular riddles, in which team work is highly needed. The agents need to learn how to cooperate in order to improve their results.
We bring into discussion the Hats riddle, the Switch riddle and the Hats* problem.
The folder results contains plots, statistics and keras models obtained from this project.
The folder other attepmts contains some unsuccessful experients.
In the Jail of Riddles there is a tradition. On each Christmas Eve the prison chief must free ten good prisoners if they are able to solve a group task in order to show that they can integrate back in the society.
This year, the chief proposes the following problem. The jailer has hats of ten different colors. He shows all the hat types to the prisoners. On each prisoner's head will be placed a hat of a random color selected from the ten colors showed earlier. The captives are placed in a circle. Thus, each prisoner can see the colors of the hats on the other nine heads.
Each prisoner must guess the color of its own hat and write it on a piece of paper. If at least one of them guesses correctly, all of them will be released. If not, all will go back to their cells.
They are not allowed to say anything or communicate in any other way with each other during the trial. However, they receive a prior one hour brake to plan together a strategy, which will prove that they are not only wise, but perfect teammates, also.
The above text was inspired in essence by problem $50 for All, which was proposed to National Security Agency (NSA) Puzzle Periodical on the month of August, 2016.