Either way, hill climbing is greedy--it'll get stuck while not exploring as effectively if as much traces as plausible are not retained. Exploration will also be more redundant if various branches do not consolidate to exploit correlations. Looked at as a general search problem
-
-
And yes,
@ReScienceEds try very hard to tackle this issue by publishing only replications of computational models (not just cognitive ones though). -
I have discussed this issue in a few different places. Here with
@NPRougier Guest, O. & Rougier, N. P. (2016). Dialogue: What is Computational Reproducibility?. IEEE CIS Newsletter on Cognitive and Developmental Systems. 13 (2). http://oliviaguest.com/doc/guest_rougier_2016.pdf … -
This blog post that
@chbergma invited me to write (thank you!): http://bootphon.blogspot.co.uk/2015/10/replication-in-computational-cognitive.html … -
This paper: Implementations are not specifications: Specification, replication and experimentation in computational cognitive modeling https://doi.org/10.1016/j.cogsys.2013.05.001 …
-
So there are more than a single issue here but I have been discussing this usually on my own and often without any support from others since the academic year of 2009/2010.
-
:( It's sad people don't see the importance of the points you've brought up. It's fascinating to see another field facing nearly the same issues as machine learning research and suggesting similar trajectories through them.
-
Yup! Not only similar issues, but I think the "root" is the same and the crossover of people/ideas from the two areas is not a coincidence!
- 2 more replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.