Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • Jul 07 01:04
    marcharper edited #1403
  • Jul 04 08:47
    drvinceknight commented #1403
  • Jul 04 08:47
    drvinceknight unlabeled #1403
  • Jul 04 08:47
    drvinceknight labeled #1403
  • Jul 04 02:19
    marcharper labeled #1403
  • Jul 03 18:41
    marcharper commented #1403
  • Jul 03 18:39
    marcharper commented #1403
  • Jul 03 18:28
    marcharper synchronize #1403
  • Jul 03 18:28

    marcharper on marcharper-patch-4

    atempt to fix typing issues unc… (compare)

  • Jul 03 18:07
    marcharper edited #1403
  • Jul 03 18:03

    marcharper on fix-dbs-documentation

    (compare)

  • Jul 03 18:03

    marcharper on run_black_on_all_files

    (compare)

  • Jul 03 18:03

    marcharper on update-contribution-documentation-for-new-branches

    (compare)

  • Jul 03 18:03

    marcharper on fix-failing-tests-from-spitefullCC

    (compare)

  • Jul 03 18:03

    marcharper on fix-failing-tests-from-spitefullCC-for-dev

    (compare)

  • Jul 03 18:03

    marcharper on add-property-based-test-for-meta-player

    (compare)

  • Jul 03 18:03

    marcharper on add-paper-by-glynatsi-et-al

    (compare)

  • Jul 03 18:03

    marcharper on bump-pyamyl-version

    (compare)

  • Jul 03 18:03

    marcharper on marcharper-patch-1-fix-type-hints

    (compare)

  • Jul 03 18:03

    marcharper on Add-JOSE-paper-by-Nik-and-Vince

    (compare)

owen (he/him)
@owen:tanti.org.uk
[m]
Hi @zewaywong
vinceknight (he/him)
@vinceknight:matrix.org
[m]
Hi @zewaywong 👋
vinceknight (he/him)
@vinceknight:matrix.org
[m]
It's not specifically required but doesn't hurt.
Thanks for opening the PR, (I've taken a look, not entirely sure why the CI has fallen over, probably an update to mypy).
owen
@owen:tanti.org.uk
[m]

Apparently, game theory is simple...

A former Conservative campaign headquarters insider concluded that No 10 was viewed as “pretty vindictive” by Tory MPs wary of scuppering their chances of promotion pretty, adding: “Once you’re in the cold you’re rarely going to come back in.” Another Tory aide said it was simple game theory. “Yes the odds of a reshuffle are slim, but are you really going to blow up your chances if you might then spend the next two, three years kicking yourself?”

https://www.theguardian.com/politics/2021/sep/07/how-boris-johnson-quelled-tory-anger-over-manifesto-breaking-tax-rise

vinceknight (he/him)
@vinceknight:matrix.org
[m]
<sigh>
vinceknight (he/him)
@vinceknight:matrix.org
[m]
I'll be doing a new Axelrod release today 👍️
vinceknight (he/him)
@vinceknight:matrix.org
[m]
New release v4.12.0
Marc
@marcharper
Neat paper on a quantum prisoner's dilemma: https://www.nature.com/articles/srep06286
While the game is derived quantum mechanically and has an entangled strategy, the game itself can be simulated classically with a modified game matrix
In this case the entangled strategy is a cooperator in the limit of no entanglement, otherwise it is essentially a kind of mixed strategy (but not the classical p S_1 + (1-p) S_2 kind)
owen
@owen:tanti.org.uk
[m]
I've started using https://docs.github.com/en/discussions elsewhere and quite like how it works - especially how it then means issues can be converted to discussions so the remaing issues become an actual todo list. Do we fancy doing that too?
Marc
@marcharper
Willing to try it out :)
vinceknight (he/him)
@vinceknight:matrix.org
[m]
Yup me too!
Also, thanks for the pointer to the quantum paper. Looks neat. Will try to read it.
Apologies for the slow reply, have been head under water for a little while (on parental leave, was planning on being more around but 🤷‍♂️.
Marc
@marcharper
No worries, hope you are enjoying leave :)
Marc
@marcharper
They cite our work:
Of particular interest, the
memory-one PSO Gambler q = [1, 0.522, 0, 0.121], an opti-
mized strategy using particle swarm algorithms in IPD games
with the conventional payoff values [18], belongs to class A
of unbending strategies (Fig. 3a), and WSLS is an unbending
strategy only if T + S < 2P (Fig. 3b).
vinceknight (he/him)
@vinceknight:matrix.org
[m]
Neat. Thanks for sharing that @marcharper
Vince Knight
@drvinceknight
Just seen this paper mentioned on twitter: https://www.nature.com/articles/s41598-022-11654-2
1 reply
They fit memory one data to histories of play (similar to our testing ZD paper: https://arxiv.org/abs/1904.00973).
owen
@owen:tanti.org.uk
[m]
Did they use our library to do it?
vinceknight (he/him)
@vinceknight:matrix.org
[m]
No although, having read only the abstract I think that's understandable.
They have data from actual human interactions and then they measure cooperation rates.
I think it would be neat to add the measure from the testing ZD paper to the library at some point so theoretically someone could look to do it.
to do it* on human interactions.
There's also scope to fit more complex behaviours like the ones we have in the library (the FSMs etc...) to historical data. That would be a neat thing to add to the library.
one day 😆
vinceknight (he/him)
@vinceknight:matrix.org
[m]
Oh neat :D
But yeah I agree.
I reached out to them on twitter and their data is available so we could always look at fitting more etc... one day
Nikoleta Glynatsi
@Nikoleta-v3
hahaha one day
"In general-sum games, the interaction of self-interested learning agents commonly leads to collectively worst-case outcomes, such as defect-defect in the iterated prisoner's dilemma (IPD). To overcome this, some methods, such as Learning with Opponent-Learning Awareness (LOLA), shape their opponents' learning process. However, these methods are myopic since only a small number of steps can be anticipated, are asymmetric since they treat other agents as naive learners, and require the use of higher-order derivatives, which are calculated through white-box access to an opponent's differentiable learning algorithm. To address these issues, we propose Model-Free Opponent Shaping (M-FOS). M-FOS learns in a meta-game in which each meta-step is an episode of the underlying ("inner") game. The meta-state consists of the inner policies, and the meta-policy produces a new inner policy to be used in the next episode. M-FOS then uses generic model-free optimisation methods to learn meta-policies that accomplish long-horizon opponent shaping. Empirically, M-FOS near-optimally exploits naive learners and other, more sophisticated algorithms from the literature. For example, to the best of our knowledge, it is the first method to learn the well-known Zero-Determinant (ZD) extortion strategy in the IPD. In the same settings, M-FOS leads to socially optimal outcomes under meta-self-play. Finally, we show that M-FOS can be scaled to high-dimensional settings. "
(we got a citation)
Nikoleta Glynatsi
@Nikoleta-v3
fyi I received an email from a current high school senior using the Axelrod Python library for one of his projects. "The project is an analysis of NIMBY interactions and I’m trying to figure out which iterative prisoners dilemma strategy is optimal in government vs residents interactions." :smile:
owen
@owen:tanti.org.uk
[m]
Is anyone here on the Python discord server? if so, you may have seen on there a request from the Python core team for library owners to test their library on 3.11 as they are trying to iron out any potential bugs. (They've done a bunch of work on the internals of CPython for 3.11).
Has anyone already tried axelrod-python out on 3.11 by any chance?
Marc
@marcharper
I haven't tried 3.11 yet, but I heard it was faster -- would be interesting to see if our tournaments are faster now
owen
@owen:tanti.org.uk
[m]
They're claiming anything between 20% and 60% faster!
owen
@owen:tanti.org.uk
[m]
Any objection to me submitting a PR to add ourselves to https://diataxis.fr/adoption/ ?
owen
@owen:tanti.org.uk
[m]
Attempting to run our test suite with 3.11 wasn't exactly a huge success. I failed to even install the dependencies!
owen
@owen:tanti.org.uk
[m]
It looks like numpy don't yet support 3.11 until it's at rc status
owen
@owen:tanti.org.uk
[m]
yep. numpy needs to be compiled from source in order to work with 3.11
Not sure I have the inclination to get into that
in other news, I've mentioned us on the Discord server recently and somebody got in touch to say thanks for showing him how to cite software libraries!
owen
@owen:tanti.org.uk
[m]
(if anyone happens to be on there, I'm meatballs#5289)