Add new strategy SelfSteem #866

janga1997 · 2017-02-17T16:52:31Z

#379
I've tried to make sure everything was done, but I might have left something. Do take a look.

janga1997 · 2017-02-17T16:55:34Z

One doubt I have is to whether to check for cases where the strategy returns a random Action.

drvinceknight · 2017-02-17T16:55:23Z

axelrod/strategies/selfsteem.py

+    If f > 0.95, 'ego' of the algorithm is inflated; always defects.
+    If 0.95 > |f| > 0.3, rational behavior; follows TitForTat algortithm.
+    If 0.3 > f > -0.3; random behavior.
+    If f < -0.95, algorithm is at rock bottom; always cooperates.


Can you add Names ... see: http://axelrod.readthedocs.io/en/latest/tutorials/contributing/strategy/writing_the_new_strategy.html This will also require you to add the reference to docs/reference/bibliography.rst.

Can you also add this strategy to docs/reference/all_strategies.rst

More info here: http://axelrod.readthedocs.io/en/latest/tutorials/contributing/strategy/adding_the_new_strategy.html

drvinceknight · 2017-02-17T16:56:54Z

axelrod/strategies/selfsteem.py

+
+    def strategy(self, opponent: Player) -> Action:
+        turns_number = len(self.history)
+        sine_value =  sin(2*pi*turns_number/10)


PEP8: 2 * pi * ...

drvinceknight · 2017-02-17T16:58:17Z

axelrod/tests/unit/test_selfesteem.py

+        self.responses_test([D], [C] * 16, [D] * 16)
+        self.responses_test([C], [D] * 9, [C] * 9)
+
+


Can you get rid of the blank lines but also include a test for the random behaviour, you can do this by passing a seed= to the responses_test function: http://axelrod.readthedocs.io/en/latest/tutorials/contributing/strategy/writing_test_for_the_new_strategy.html

What exactly is happening here with seed ?
I saw in seed =1 and seed = 2 being passed. What does it do?

So a random process starts with a random seed. This just ensures the "randomness" can be repeated:

>>> import random >>> random.seed(0) >>> random.random() 0.8444218515250481 >>> random.seed(0) 0.8444218515250481

The seed=1 and seed=2 just set those seeds for the strategies.

janga1997 · 2017-02-17T18:39:29Z

@drvinceknight Does the build error have something to do with my PR?
I saw a lot of errors about documentation and moran.py

drvinceknight · 2017-02-17T18:45:29Z

docs/reference/all_strategies.rst

@@ -135,6 +135,9 @@ Here are the docstrings of all the strategies in the library.
 .. automodule:: axelrod.strategies.shortmem
   :members:
   :undoc-members:
+.. automodule:: axelrod.strategies.selfsteem
+  :members:


Your indentation levels aren't correct here. Compare them to the other strategies.

drvinceknight · 2017-02-17T18:47:30Z

@drvinceknight Does the build error have something to do with my PR?
I saw a lot of errors about documentation and moran.py

The error is in the building of the docs:

Warning, treated as error:
/home/travis/build/Axelrod-Python/Axelrod/axelrod/strategies/selfsteem.py:docstring of axelrod.strategies.selfsteem.SelfSteem:5: ERROR: Undefined substitution referenced: "f".

It might be due to the indentation I pointed out but it could be something else. If you run make html in the /docs dir you can make the docs locally and see what the error is (you might need to pip install a couple of sphinx dependencies).

drvinceknight · 2017-02-17T18:51:23Z

There are other errors, I'm taking a look now.

drvinceknight · 2017-02-17T18:59:03Z

axelrod/strategies/selfsteem.py

+    with the current iteration.
+
+    If f > 0.95, 'ego' of the algorithm is inflated; always defects.
+    If 0.95 > |f| > 0.3, rational behavior; follows TitForTat algortithm.


This line is creating an error as sphinx is confused by |f|. I suggest replacing it with abs(f).

drvinceknight · 2017-02-17T18:59:33Z

axelrod/strategies/selfsteem.py

+            return opponent.history[-1]
+
+        elif sine_value < 0.3 and sine_value > -0.3:
+            return random.choice((C, D))


Better to use the inbuilt random_choice() from axelrod.random_.

drvinceknight · 2017-02-17T18:59:59Z

axelrod/strategies/selfsteem.py

+    - SelfSteem: [Andre2013]_
+    """
+
+    name = 'SelfSteem'


What is the name of this strategy?

SelfSteem or SelfEsteem? Across module names etc you seem to be combining these.

It is SelfSteem. I mistakenly named the test test_selfesteem instead of test_selfsteem.

drvinceknight · 2017-02-17T19:01:04Z

My suggestions there fix some of the errors but not all of them. My hunch is that it's something offsetting the random seed somewhere but I'd need more time to look through it. I'll try and get to it over the weekend if you don't find the fix but I'm travelling so it might be delayed :)

I'm sure we'll figure it out :)

drvinceknight · 2017-02-17T19:03:50Z

axelrod/tests/unit/test_selfesteem.py

+        self.responses_test([D], [C] * 16, [D] * 16)
+        self.responses_test([C], [D] * 9, [C] * 9)
+
+


So a random process starts with a random seed. This just ensures the "randomness" can be repeated:

>>> import random >>> random.seed(0) >>> random.random() 0.8444218515250481 >>> random.seed(0) 0.8444218515250481

The seed=1 and seed=2 just set those seeds for the strategies.

Fix the classification and also add a test for stochastic behaviour.

drvinceknight · 2017-02-18T14:27:51Z

@janga1997 I've opened a PR on your fork: janga1997#1

If you merge that in, it will automatically update the PR with my suggested commits.

The main error you were getting was caused because of a mis classification: you were saying the strategy was not stochastic and so it was being used in other tests (but it is stochastic).

Fix strategy.

janga1997 · 2017-02-18T15:35:59Z

@drvinceknight I've been seeing this error since the beginning.

File "./docs/tutorials/advanced/classification_of_strategies.rst", line 49, in classification_of_strategies.rst

Failed example:

    len(strategies)

Expected:

    64

Got:

    65

janga1997 · 2017-02-18T15:40:40Z

@drvinceknight And why did we merge #839 ?
It seems to throw a lot of errors on mypy.

drvinceknight · 2017-02-18T15:42:08Z

@drvinceknight I've been seeing this error since the beginning.

Yup, that will be the doctest for this page: http://axelrod.readthedocs.io/en/latest/tutorials/advanced/classification_of_strategies.html

Just bump the number.

drvinceknight · 2017-02-18T15:42:49Z

@drvinceknight And why did we merge #839 ?
It seems to throw a lot of errors on mypy.

You're completely correct! I'm not sure how that got by. I'm looking at it now. I might revert it.

… SelfSteem

janga1997 · 2017-02-18T16:04:09Z

@drvinceknight I guess that was the main error. The tests seem to be passing now.

drvinceknight · 2017-02-18T16:13:51Z

Yup, the only thing left is those errors about the type annotations in the moran process and as I've written on #868 that's not being picked up by travis :)

The strategy looks good 👍 I noted there's one point in the paper:

Furthermore, the algorithms implements a retaliation policy, if the opponent defects; the sin curve is shifted (changing the sin phase).

As far as I can tell there's no indication as to how that is done right?

Could you also tweak in docs/reference/bibliography.rst and move [Andre2013] ... to be in alphabetical order?

janga1997 · 2017-02-18T16:22:58Z

@drvinceknight It didn't specify anything, not even how much the phase change is supposed to be, and in which direction.

drvinceknight · 2017-02-18T19:04:36Z

@drvinceknight It didn't specify anything, not even how much the phase change is supposed to be, and in which direction.

Yeah I see that. Not sure if this means it's worth rewording the docstring to make a note of this. Something like "This strategy is based on ..., however due to a limited number of details in the original description the sin function is not shifted as in the original work." Any thoughts @marcharper?

I'm happy with everything else.

marcharper · 2017-02-22T05:32:31Z

I agree with @drvinceknight -- let's make a note in a docstring and email the authors for a clarification. Unfortunately this is a common issue and one of the goals of the library is to enable reproducible results going forward. It's clear that journals are not requiring source code disclosures and peer reviewers are not demanding sufficient descriptions :/

marcharper · 2017-02-22T05:33:14Z

axelrod/strategies/selfsteem.py

+
+class SelfSteem(Player):
+    """
+    This strategy is based on the feeling with the same name.


self esteem?

It means self esteem, that's how its named in the paper. I decided not to change it.

marcharper · 2017-02-22T05:33:33Z

axelrod/strategies/selfsteem.py

+    """
+    This strategy is based on the feeling with the same name.
+    It is modeled on the sine curve(f = sin( 2* pi * n / 10 )), which varies
+    with the current iteration.


current iteration n is the round number? (I see below the answer is yes but the docstring could say so.)

marcharper · 2017-02-22T05:35:24Z

axelrod/strategies/selfsteem.py

+
+    def strategy(self, opponent: Player) -> Action:
+        turns_number = len(self.history)
+        sine_value =  sin(2 * pi * turns_number / 10)


Just checking -- is this integer or float division in the reference?

In python3, doesn't it return a float by default?

It does, I'm just making sure that's correct.

Yes, we do need a float division.

janga1997 · 2017-02-22T05:40:18Z

@marcharper I've emailed the authors of the paper regarding what I'm trying to do here, and what information I require. Yet to receive a reply.

marcharper · 2017-02-22T06:01:03Z

Great, let's give them a day or two to reply. If we don't hear anything then we'll go with our best guess.

janga1997 added 3 commits February 17, 2017 22:16

Add new stratefy SelfSteem

57a94d7

Add tests for SelfSteem

b58b6b9

Add strategy to _strategies.py

15463da

drvinceknight requested changes Feb 17, 2017

View reviewed changes

Changes to documentation and style

21ad4d7

drvinceknight requested changes Feb 17, 2017

View reviewed changes

janga1997 and others added 3 commits February 18, 2017 00:49

Style and naming

a50750f

Use random_choice()

f7654bc

Fix strategy.

04ed431

Fix the classification and also add a test for stochastic behaviour.

Merge pull request #1 from Axelrod-Python/janga1997-SelfSteem

d00e419

Fix strategy.

drvinceknight mentioned this pull request Feb 18, 2017

Revert "Add type for Moran" #867

Merged

janga1997 added 2 commits February 18, 2017 21:18

Edit to classification_of_strategies

6623e00

Merge branch 'SelfSteem' of https://github.com/janga1997/Axelrod into…

aacf48e

… SelfSteem

Rearrange bibliography.rst

09a94eb

drvinceknight approved these changes Feb 18, 2017

View reviewed changes

drvinceknight mentioned this pull request Feb 19, 2017

Add new strategy VeryBad #869

Merged

meatballs added the ready-to-merge label Feb 19, 2017

marcharper approved these changes Feb 22, 2017

View reviewed changes

Update docstrings to reflect lack of information

90b8bd8

meatballs merged commit daf0c1d into Axelrod-Python:master Feb 26, 2017

janga1997 changed the title ~~Add new Strategy SelfSteem~~ Add new strategy SelfSteem Mar 11, 2017

		self.responses_test([D], [C] * 16, [D] * 16)
		self.responses_test([C], [D] * 9, [C] * 9)

Add new strategy SelfSteem #866

Add new strategy SelfSteem #866

Conversation

janga1997 commented Feb 17, 2017

janga1997 commented Feb 17, 2017 • edited Loading

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

janga1997 commented Feb 17, 2017

Choose a reason for hiding this comment

drvinceknight commented Feb 17, 2017

drvinceknight commented Feb 17, 2017

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

drvinceknight commented Feb 17, 2017

Choose a reason for hiding this comment

drvinceknight commented Feb 18, 2017

janga1997 commented Feb 18, 2017

janga1997 commented Feb 18, 2017

drvinceknight commented Feb 18, 2017

drvinceknight commented Feb 18, 2017

janga1997 commented Feb 18, 2017

drvinceknight commented Feb 18, 2017

janga1997 commented Feb 18, 2017

drvinceknight commented Feb 18, 2017 • edited Loading

marcharper commented Feb 22, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

marcharper Feb 22, 2017 • edited Loading

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

janga1997 commented Feb 22, 2017

marcharper commented Feb 22, 2017

janga1997 commented Feb 17, 2017 •

edited

Loading

drvinceknight commented Feb 18, 2017 •

edited

Loading

marcharper Feb 22, 2017 •

edited

Loading