[RFR] Get board reward #2

floo51 · 2016-09-22T13:32:30Z

⚠️ Waiting for merge + rebase : do not review

Get board reward
Add tests

floo51 · 2016-09-22T14:21:53Z

Rebased, ready for review

Kmaschta · 2016-09-22T14:33:05Z

ai/reward.py

+player_has_consecutive_chunks = tf.equal(consecutive_chunks, 1)
+is_game_won = tf.reduce_any(player_has_consecutive_chunks)
+
+get_board_reward = tf.case(


Make it more readable:

get_board_reward = tf.case( {is_game_won: reward_positive, is_game_lost: reward_negative}, default=reward_none, exclusive=True )

Kmaschta · 2016-09-22T14:35:01Z

connectfour.py

-    print session.run(is_aligned, feed_dict={
-        board: game_board,
-        mask: bitmasks[2]
+    print session.run(get_board_reward, feed_dict={


In python 3, print is not a built-in keyword anymore but a function.
Please use it as a function to avoid error when you'll switch to python 3.

Kmaschta · 2016-09-22T14:35:58Z

Can you hook unit tests with Travis please?

…e call

floo51 · 2016-10-03T07:10:20Z

Last version pushed, as per demo.
Previous comments taken into account, ready for new review.

Kmaschta

Please use one of these format instead of string concatenations:

>>> a, b  = 5, 10
>>> 'some {} thing {}'.format(a, b)
"some 5 thing 10"
>>> 'foo %s bar %s' % (a, b)
"foo 5 bar 10"

Great work!

Kmaschta · 2016-10-03T12:18:46Z

ai/human/human.py

+
+    def turn_feedback(self, player, column):
+        who = 'You'
+        if self.player != player:


who = 'You' if self.playeur == playeur else 'Opponent'

Kmaschta · 2016-10-03T12:24:16Z

ai/minimax/minimax.py

+                game_copy.play(i, current_player)
+                legal_moves.append(game_copy)
+
+        if depth == 0 or len(legal_moves) == 0 or game.get_status() == GAME_STATUS['FINISHED']:


Give a name to your conditions in order to make it more readable. For example:

first_turn = depth == 0 no_legal_moves = len(legal_moves) == 0 game_finished = game.get_status() == GAME_STATUS['FINISHED'] if first_turn or no_legal_moves or game_finished: pass # Do the flop

Kmaschta · 2016-10-03T12:24:47Z

ai/minimax/minimax.py

+
+        if opp_fours > 0:
+            return -100000
+        else:


Useless else statement

Kmaschta · 2016-10-03T12:24:58Z

ai/minimax/test_minimax.py

@@ -0,0 +1,5 @@
+# import unittest


Commented out code

Kmaschta · 2016-10-03T12:27:30Z

connectfour/game.py

+from tool.array_transposer import transpose_horizontally, transpose_diagonally
+
+
+GAME_STATUS = {


Not important for this review, but the good practice in python is to define constants list as:

GAME_PLAYING = 0 GAME_FINISHED = 1 GAME_STATUS = ( ('PLAYING', GAME_PLAYING), ('FINISHED', GAME_FINISHED), )

Kmaschta · 2016-10-03T12:30:44Z

connectfour/game.py

+        elif player2_won:
+            self.winner = 1
+
+        status = GAME_STATUS['FINISHED'] \


Where the ternary condition is too long, it's better to use the classical if/else condition

Kmaschta · 2016-10-03T12:31:11Z

connectfour/game.py

+        aligned_discs = self.has_aligned_discs(number_of_cells)
+        if player in aligned_discs:
+            return aligned_discs[player]
+        else:


Useless else

Florian FERBACH added 2 commits September 22, 2016 16:14

Add board reward function

879ba6d

Fix test imports

deebe1e

floo51 force-pushed the get_board_reward branch from dce48df to deebe1e Compare September 22, 2016 14:18

floo51 changed the title ~~[WIP] Get board reward~~ [RFR] Get board reward Sep 22, 2016

Kmaschta suggested changes Sep 22, 2016

View reviewed changes

Florian FERBACH added 10 commits September 26, 2016 16:22

Add minimax algorithm

08bd5f9

Move neural network ai into its own folder

763a902

Separate different trainings

fe7bfb7

Add minimax algorithm

b034e56

Reorder algorithms so they can be chosen dynamically from command lin…

6e9be47

…e call

Change from manual to automatic back propagation

dcc380c

Move array transpose and array consecutive logic outside minimax

b952e15

Move game logic into its own file

a92da34

Add human AI

8bf1a98

Amend Makefile to fit main file changes

d80cde7

floo51 mentioned this pull request Oct 3, 2016

Makefile install #3

Open

Kmaschta suggested changes Oct 3, 2016

View reviewed changes

fzaninotto merged commit 9a6f427 into master Oct 7, 2016

fzaninotto deleted the get_board_reward branch October 7, 2016 12:38

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[RFR] Get board reward #2

[RFR] Get board reward #2

floo51 commented Sep 22, 2016 •

edited

floo51 commented Sep 22, 2016

Kmaschta Sep 22, 2016

Kmaschta Sep 22, 2016

Kmaschta commented Sep 22, 2016

floo51 commented Oct 3, 2016

Kmaschta left a comment

Kmaschta Oct 3, 2016

Kmaschta Oct 3, 2016

Kmaschta Oct 3, 2016

Kmaschta Oct 3, 2016

Kmaschta Oct 3, 2016

Kmaschta Oct 3, 2016

Kmaschta Oct 3, 2016

		from tool.array_transposer import transpose_horizontally, transpose_diagonally


		GAME_STATUS = {

[RFR] Get board reward #2

[RFR] Get board reward #2

Conversation

floo51 commented Sep 22, 2016 • edited

floo51 commented Sep 22, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Kmaschta commented Sep 22, 2016

floo51 commented Oct 3, 2016

Kmaschta left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

floo51 commented Sep 22, 2016 •

edited