-
Notifications
You must be signed in to change notification settings - Fork 1
/
README
35 lines (30 loc) · 2.15 KB
/
README
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
--------------------------------------------------------------------------------
| /$$ /$$ /$$ |
| | $$ | $$ |__/ |
| | $$ /$$ /$$ /$$$$$$$| $$ /$$ /$$ |
| | $$ | $$ | $$ /$$_____/| $$ /$$/| $$ |
| | $$ | $$ | $$| $$ | $$$$$$/ | $$ |
| | $$ | $$ | $$| $$ | $$_ $$ | $$ |
| | $$$$$$$$| $$$$$$/| $$$$$$$| $$ \ $$| $$ |
| |________/ \______/ \_______/|__/ \__/|__/ |
| |
--------------------------------------------------------------------------------
| Written by Thomas Schaffer <thomas.schaffer@epitech.eu> in Java. |
| Licensed under the MIT license. |
--------------------------------------------------------------------------------
--------------------------------------------------------------------------------
| Description |
--------------------------------------------------------------------------------
Lucki is a Poker AI agent that I developed for my MSc dissertation.
It is based on some research by the Poker research group at the University
of Alberta, notably their bots Poki, Loki, etc.
The global architecture of the agent (hand weight tables for the opponents,
and an hand evaluator enumerating potential future draws and applying
those weights) is inspired by their published research and completely
reimplemented from scratch.
This project's contribution was to try and improve on the somewhat limited
Betting Strategy component by experimenting with neural networks trained
by a genetic algorithm.
This contains the agent, along with a complete framework for poker play
and AI agents so that it should be easy to implement other strategies
while benefiting from the existing hand evaluation / profiling components.