Updated Huston Cohen 1994 model #2222
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Set back all the parameter values to the original settings as in the 1994 Cohen and Huston paper. Most importantly: now the weights before and after initialization are set correctly. I've also changed it so neutral input can be 1 (weight for neutral input was on 0), and I've changed it so that both word/color --> response and response --> word/color are changed before and after initialization. No idea what the rationale was from the original author how it was done, but for me it didn't make sense!
Comma missing and changed settle time to 500 (does not matter qualitatively, but looks more like the cohen/huston 1994 plot). And made some things prettier.
|
This pull request fixes 4 alerts when merging 0788f44 into 912f691 - view on LGTM.com fixed alerts:
|
|
This PR causes the following changes to the html docs (ubuntu-latest-3.7-x64): See CI logs for the full diff. |
|
Here's a comparison for the figures produced: devel Fig 2: devel I don't know a correct comparison for Fig 1, but Fig 2 is in the original paper: While it looks to me that the reaction times in Fig 2 devel match the paper more closely, I'm not sure if this was part of the motivation. Thoughts on this @jdcpni @tylergiallanza? |
|
This pull request fixes 4 alerts when merging 0788f44 into a1d4f2a - view on LGTM.com fixed alerts:
|
|
The motivation to change the script was because the weights before and after the initialization run were not set correctly! Thus, the first graphs are produced with the weights between colour/word and response on 0 for both initialization and actual runs. That is by design incorrect, thus the graphs can't really be compared in that regard. |
Fair, but this suggests to me that there are additional errors somewhere. If the script is meant to replicate the original model, then it should match the reaction times. It seems that either the original simulation was also incorrect in the same way, though this seems unlikely in part due to its matching the empirical data, or that there are additional changes that need to be made to the script with weights set for post-initialization runs to bring the results in line with the paper. |
|
In the original Psyneulink paper, to get to the results of the 1994 model, some changes needed to be made to several key parameters deviating from the 1994 model:
Changing that back to the 1994 parameter setting in the original Psyneulink script leads to a never-ending simulation since the threshold can't be reached. In the new script I propose I'm able to keep the 1994 parameter values, whilst replicating the 1994 results. Not exactly though, since this script runs a single noiseless simulation, whilst the 1994 results show a noise-infused average of 100 trials. |
|
Do we have results in the corrected PNL version running it with noise for 100 trials? (Tyler, cc’ing you as I’m not sure you are receiving messages in this thread).
jdc
On Dec 7, 2021, at 3:21 AM, JeshuaT ***@***.******@***.***>> wrote:
In the original Psyneulink paper, to get to the results of the 1994 model, some changes needed to be made to several key parameters deviating from the 1994 model:
* integration rate from 0.01 to 0.1
* threshold from 0.6 to 0.55
Changing that back to the 1994 parameter setting in the original Psyneulink script leads to a never-ending simulation since the threshold can't be reached. In the new script I propose I'm able to keep the 1994 parameter values, whilst replicating the 1994 results. Not exactly though, since this script runs a single noiseless simulation, whilst the 1994 results show a noise-infused average of 100 trials.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub<#2222 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ACSHKGBVG4DOW52EFJO6DGTUPW7XJANCNFSM5JJWBIYA>.
Triage notifications on the go with GitHub Mobile for iOS<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android<https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Set back all the parameter values to the original settings as in the 1994 Cohen and Huston paper. Most importantly: now the weights before and after initialization are set correctly. I've also changed it so neutral input can be 1 (weight for neutral input was on 0), and I've changed it so that both word/color --> response and response --> word/color are changed before and after initialization. No idea what the rationale was from the original author how it was done, but for me, it didn't make sense!
Furthermore, settling time to 500, and added variables slope and intercept. Otherwise, I've tried to leave it as much as how the original model was written.