Moran process #534

marcharper · 2016-04-13T06:21:02Z

Closes #224

The Moran process is somewhat of a hybrid of a match and a tournament. It's sort of a n >= 2 match, but also has several rounds. The process is widely studied in the literature.

Doc tests to skip Doctest Add Moran doc to toc tree

Fix tests

drvinceknight · 2016-04-13T07:27:27Z

axelrod/moran.py

+
+    def play(self):
+        """Play the process out to completion."""
+        while True: # O_o


Why this inline comment?

It's not necessary

drvinceknight · 2016-04-13T07:53:00Z

This looks really great @marcharper, I'm heading to class now. Going to take another look at it later.

Docstring improvement More review updates

meatballs · 2016-04-13T09:17:43Z

axelrod/moran.py

+        if self.noise:
+            return True
+        else:
+            return any(p.classifier['stochastic'] for p in self.players)


This method is almost identical to the one in the Match class. I think they both ought to be abstracted out somewhere.

drvinceknight · 2016-04-13T09:55:20Z

@marcharper the winner method returns the str of the player. This is in contrast to the Match class which returns the actual winning player instance. I understand why this is (this is an evolutionary process so an actual winner is not a particular player but a strategy). Could we change the name of this method to something that clarifies that:

winning_strategy_name
winner_name

? Not too sure what but I don't like that the type of Match.winner() and Moran.winner() is different.

drvinceknight · 2016-04-13T09:56:26Z

One of my comments on the docs seemed to have been described as 'on an outdated diff':

On further_topics/probabilistict_end_tournaments.rst we have:

>>> results = tournament.play()
>>> m = results.payoff_matrix
>>> for row in m:  
...     print([round(ele, 1) for ele in row]) # Rounding output

Perhaps do that in the docs here for consistency?

drvinceknight · 2016-04-13T15:21:35Z

docs/tutorials/getting_started/moran.rst

+The sequence of populations::
+
+    >>> import pprint
+    >>> pprint.pprint(mp.populations)  # doctest: +SKIP


Given the base print below, do we want the same here for consistency? (minor)

drvinceknight · 2016-04-13T15:22:45Z

A part from my pprint/print comment inline my one remaining comment about the docs would be whether it should sit in getting_started or if it was better suited further_topics where for example the probabilistic ending tournament sits?

marcharper · 2016-04-13T15:39:08Z

I think pprint does a better job in less code in this case -- I don't have an opinion on the docs, whatever you think is best is fine.

drvinceknight · 2016-04-13T15:39:37Z

I think pprint does a better job in less code in this case.

Fine.

drvinceknight · 2016-04-13T15:41:29Z

whatever you think is best is fine.

Can you move it to further_topics then? It feels better there to me. I would put it between Morality Metrics and Probabilistic Ending Tournaments.

More review updates more review

marcharper · 2016-04-13T15:43:47Z

Ok, but I'd say the same logic applies to the ecological variant, no?

drvinceknight · 2016-04-13T15:46:36Z

Ok, but I'd say the same logic applies to the ecological variant, no?

I think that given that it comes out of the standard tournament/results class (it's a simple manipulation of the payoff matrix) I'd suggest that it should stay there. The moran process is as you said a different type of interaction all together...

Fix doc test

marcharper · 2016-04-13T19:56:24Z

I've moved it but I don't agree with your position. IMO the main value of the library comes from

the library of strategies
the test suite
the classes that run matches and tournaments

The ecological variant is somewhat of a side curiosity that isn't going to excite anyone about the library; the Moran process and other population dynamics that are actively studied would actually be of interest to other researchers. Just because the ecological variant falls out of the results for the tournament code doesn't really make it a first class citizen. I could code the Moran process to use the tournament objects to run its matches and achieve the same effect.

drvinceknight · 2016-04-13T20:15:38Z

I've moved it but I don't agree with your position. IMO the main value of the library comes from

Thanks :) Have marked this with a positive review.

the library of strategies
the test suite
the classes that run matches and tournaments

I completely agree. I would suggest a 4th strength is the accessibility of the library: a few people have been introduced to dev, game theory and open source through it (something I'm completely delighted about).

The ecological variant is somewhat of a side curiosity that isn't going to excite anyone about the library; the Moran process and other population dynamics that are actively studied would actually be of interest to other researchers. Just because the ecological variant falls out of the results for the tournament code doesn't really make it a first class citizen. I could code the Moran process to use the tournament objects to run its matches and achieve the same effect.

I see what you're saying and understand/agree that from a high level game theoretic research point of view (which is certainly what the library is) that the moran process is far more of interest. However, I think the getting started tutorial should perhaps be aimed also at people who have very little knowledge of game theory. Explaining that the eco variant is based on the payoff matrix is simpler and more immediately accessible than a different interaction type: as you said the moran process is a hybrid of an n player match and a tournament.

I do not at all mean that the moran process is a second citizen (far from it, and apologies if that is how I came across with this awesome contribution of yours), I just don't feel it's what's needed in the introduction (which is not only aimed at experts in game theory but also complete novices). I am 98% sure that with some of my students for whom I've said "work through the intro", I would have probably said "except the moran process", only because it requires them to create a different class to the Tournament, Results, Plot (some of my students have come to the library with not the strongest code/game theory background).

No doubt experts/researchers would perhaps even come looking for the moran process and I don't think it's too hard to find (just as for the probabilistic ending tournaments).

Anyway, sorry, I realise you were trying to put this to bed (perhaps we can open issue if we should chat about this further: onboarding is important to get right). I just wanted to make sure you didn't think I thought this was a second class citizen. I really do think it's awesome and am excited to learn more about it. The fact that it's heavily used in the literature is awesome. :) 👍

drvinceknight · 2016-04-13T20:25:53Z

Tournament, Results, Plot (some of my students have come to the library with not the strongest code/game theory background).

Having written that and realised that we need the Eco class...

Perhaps moving the ecological variant to the Further topics is the way to go. I'll leave it up to you :), either of these is fine by me:

As is.
Both ecological and Moran in further (perhaps this further simplifies the intro which is even better).

marcharper · 2016-04-13T21:28:49Z

Let's discuss in a future PR / on gitter. I agree that accessibility is both important and valuable aspect of the library.

I think this one is good to go in pending a second review.

langner · 2016-04-14T05:36:52Z

Good stuff!

My opinion about the docs: create something like beyond_basics.rst which mentions both moran and eco and links to pages that expound on those topics. I would also put noisy tournaments in that bucket. In my mind, getting started should contain only the absolute basics; it should be geared towards people that have just learned about IPD yesterday, maybe this is also the first they've seen the words "game" and "theory" next to each other.

drvinceknight · 2016-04-14T07:43:11Z

@meatballs this is good to go I believe (assuming you're happy). The conversation about the docs has been moved to #536 :)

Moran process and tests

5269133

marcharper force-pushed the moran-process branch from 89600a3 to ed156f0 Compare April 13, 2016 06:38

marcharper added 2 commits April 13, 2016 00:22

Docs for Moran Process

9a3494d

Doc tests to skip Doctest Add Moran doc to toc tree

Fix random differences between python 2 and 3

9de867e

Fix tests

marcharper force-pushed the moran-process branch from 28f2ef4 to 9de867e Compare April 13, 2016 07:22

drvinceknight mentioned this pull request Apr 13, 2016

Pickle Protocol Version #530

Closed

drvinceknight reviewed Apr 13, 2016
View reviewed changes

Review fixes

fc831f3

Docstring improvement More review updates

marcharper force-pushed the moran-process branch from 9a2d9e5 to fc831f3 Compare April 13, 2016 08:01

meatballs reviewed Apr 13, 2016
View reviewed changes

drvinceknight added 2 commits April 13, 2016 10:45

Adding hypothesis test and specific examples.

1146f3a

Correcting rst syntax.

5b28eec

drvinceknight reviewed Apr 13, 2016
View reviewed changes

Abstract is_stochastic from Match and Moran

b487bde

More review updates more review

marcharper force-pushed the moran-process branch from 697be81 to b487bde Compare April 13, 2016 15:41

Fix Moran Tests

9a950d8

Fix doc test

marcharper force-pushed the moran-process branch from 609a07a to 9a950d8 Compare April 13, 2016 19:42

drvinceknight added the 1-positive-review label Apr 13, 2016

marcharper mentioned this pull request Apr 14, 2016

Docs Discussion #536

Closed

meatballs merged commit 737fd4e into Axelrod-Python:master Apr 14, 2016

marcharper deleted the moran-process branch April 16, 2016 15:31

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Moran process #534

Moran process #534

marcharper commented Apr 13, 2016

drvinceknight Apr 13, 2016

marcharper Apr 13, 2016

drvinceknight commented Apr 13, 2016

meatballs Apr 13, 2016

drvinceknight commented Apr 13, 2016

drvinceknight commented Apr 13, 2016

drvinceknight Apr 13, 2016

drvinceknight commented Apr 13, 2016

marcharper commented Apr 13, 2016

drvinceknight commented Apr 13, 2016

drvinceknight commented Apr 13, 2016

marcharper commented Apr 13, 2016

drvinceknight commented Apr 13, 2016

marcharper commented Apr 13, 2016

drvinceknight commented Apr 13, 2016

drvinceknight commented Apr 13, 2016

marcharper commented Apr 13, 2016

langner commented Apr 14, 2016

drvinceknight commented Apr 14, 2016

Moran process #534

Moran process #534

Conversation

marcharper commented Apr 13, 2016

drvinceknight Apr 13, 2016

Choose a reason for hiding this comment

marcharper Apr 13, 2016

Choose a reason for hiding this comment

drvinceknight commented Apr 13, 2016

meatballs Apr 13, 2016

Choose a reason for hiding this comment

drvinceknight commented Apr 13, 2016

drvinceknight commented Apr 13, 2016

drvinceknight Apr 13, 2016

Choose a reason for hiding this comment

drvinceknight commented Apr 13, 2016

marcharper commented Apr 13, 2016

drvinceknight commented Apr 13, 2016

drvinceknight commented Apr 13, 2016

marcharper commented Apr 13, 2016

drvinceknight commented Apr 13, 2016

marcharper commented Apr 13, 2016

drvinceknight commented Apr 13, 2016

drvinceknight commented Apr 13, 2016

marcharper commented Apr 13, 2016

langner commented Apr 14, 2016

drvinceknight commented Apr 14, 2016