Tasks.todo

This file is formatted for the `plaintasks` plugin (lots of love to you!) for Sublime Text 3.
It is in markdown and any markdown editor should in principle be able to read/edit this file.
Development:
 ☐ Document configuration options specified in config.py
 ✔ ArgParser @done (17-03-31 14:15)
 ✔ Snake with increasing length @done (17-03-30 02:42)
 ✔ Head differently colored @done (17-03-30 02:42)
 ✔ Prevent reverse moves @done (17-03-30 02:42)
 ✔ Fix toggling between GUI and non GUI versions @done (17-04-01 14:54)
 ✔ Move score, foodScore and livingScore to gameState @done (17-03-30 02:42)
 ✔ Check if food is valid @done (17-03-30 02:43)
 ✔ Check if update causes death @done (17-03-30 02:43)
 ☐ Special power food
 ✔ Add border to cells @done (17-04-01 14:54)
 ✔ Add gameover message @done (17-03-30 02:43)
 ☐ Add walls
 ☐ Add some graphics which actually looks like head/tail
 ☐ Typo; replace all 'cordinate' with 'coordinates'
 ☐ Whenever a next action or related method is being called, check first if game is over!
 ☐ Add a proper way to specify the dependency between certain search agents and the depth argument
 ✔ PRIORITY: Modify `dlsnake.base.gameState` to use lazy evaluation for grid creation. @done (17-04-02 01:05)
 Technically, grid construction is unnecessary for the logic to work 
 and this lazy evaluation can significantly speed things up.
 ☐ `--csv` and `--silent` options seem to be broken. Investigate.
 ☐ Add flag to save pygame rendered images to folder.
 ✔ Create a separate script to simulate on multiple threads. @done (17-04-05 21:35)
 ☐ Figure out how to dump and load trained data and game episodes.

Pertaining to Project:
  ✔ Implement manual play @done (17-03-30 02:43)
  ✔ Implement Reflex Agent @done (17-03-30 02:43)
  ☐ Implement intelligent adversary controlling food capsules so as to trap snake.
    ✔ Adversary maximizing Manhattan distance from snake @done (17-03-30 04:06)
    ☐ Adversary trying to place food in circular configurations formed by snake
    ☐ Adversary trying to place food near walls
  ✔ Implement adversarial search (minMax) @done (17-04-01 14:54)
    ✔ Verify the algorithm in depth @done (17-04-01 14:55)
     There is a lot of nuts and bolts that can be loose. Analyze the algorithm
     especially since your hypothesis of minMax trees growing out of proportion
     turned out to be wrong.
     ☐ Implement Alpha-Beta Pruning
     ☐ Implement a variation wherein the living penalty and food scores are the only rewards.
     That is, replace the evaluation function used at the terminal state with 0 and instead
     assign values to actions. All non-food eating action gets -1 and all food eating actions
     get +50.
  ✔ Implement expectimax search @done (17-04-05 21:33)
  Won't do.
  ✔ Implement MDP based Q-Learning. Won't do @done (17-04-05 21:33)
  ☐ Implement Feature based approximate Q-learning
  ☐ Extract features from deep learning.