Start v3 implemention #18
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
Adding new SYN stage (synape model) for a v3 CARFAC.
|
Rob says I added files to the wrong dir, not under matlab. I don't see how to move them, or how to delete/cancel/abandon this pull request and try again, but I'll work on it. |
robsc
reviewed
Sep 9, 2024
matlab/CARFAC_Test.m
Outdated
| switch test_freqs(k) | ||
| case 300 | ||
| expected_results = [ ... | ||
| [1.898725, 450.438353]; |
There was a problem hiding this comment.
These goldens look quite different to the 2 cap - which i guess is expected but I don't know how to "know" they're good/bad.
There was a problem hiding this comment.
I'm not thinking the numbers are good yet. They are just what makes it pass for now. And when I take out that fs = 48000 they're probably wrong anyway. So treat these as placeholders.
Fixed Design process per robsc's comments on PR 18.
Some SYN improvements, but really it needs a revamp of the Design process, with better parameterization and automatic calculation of rest rate to subtract to get AGC back closer to how it worked in old versions.
I reparameterized the SYN stage in terms of spontaneous and saturation rates for the classes and few other things, and implemented the design process to account for everything to get those to come out right, all in terms of variables that show up on the diagram that's an augmentation of the v2 two_cap IHC diagram (for documentation, in progress). Still need to move some of the numbers into SYN_params.
robsc
approved these changes
Sep 23, 2024
.DS_Store
Outdated
There was a problem hiding this comment.
This file is a macos fragment and I don't think it belongs here.
Moving a bunch of magic numbers out of the SYN_Design code and into the default SYN_params, so they are modifiable by user code. And various other code tweaks and naming improvements (e.g. changed z to r for release rate; short variable names to facilitate labeling on doc diagrams).
The agc_weights needed to be appropriately set up to do the right thing with different sample rates, and to allow different numbers of IHCs per channel. The AGC feedback is still proportional to the number of fibers, healthy or not, so the agc_weights parameters might need adjustment if the healthy_n_fibers is changed. The param n_fibers was renamed healthy_n_fibers, as that's what applied at design time, and as the n_fibers in the coeffs may change with model training (in the JAX version at least).
copybara-service bot
pushed a commit
that referenced
this pull request
Oct 2, 2024
-- 628c358 by Richard Lyon <[email protected]>: Start v3 implemention Adding new SYN stage (synape model) for a v3 CARFAC. -- 3ab484e by Richard Lyon <[email protected]>: Add files via upload -- ab450a0 by Richard Lyon <[email protected]>: Fixes per robsc's comments on PR 18 Fixed Design process per robsc's comments on PR 18. -- 6975bde by Richard Lyon <[email protected]>: Some SYN improvements; still needs a lot of work. Some SYN improvements, but really it needs a revamp of the Design process, with better parameterization and automatic calculation of rest rate to subtract to get AGC back closer to how it worked in old versions. -- 17f5564 by Richard Lyon <[email protected]>: Reparameterized, better design phase I reparameterized the SYN stage in terms of spontaneous and saturation rates for the classes and few other things, and implemented the design process to account for everything to get those to come out right, all in terms of variables that show up on the diagram that's an augmentation of the v2 two_cap IHC diagram (for documentation, in progress). Still need to move some of the numbers into SYN_params. -- b79778a by Richard Lyon <[email protected]>: Move SYN constants from code into default parameters Moving a bunch of magic numbers out of the SYN_Design code and into the default SYN_params, so they are modifiable by user code. And various other code tweaks and naming improvements (e.g. changed z to r for release rate; short variable names to facilitate labeling on doc diagrams). -- 0181d1e by Richard Lyon <[email protected]>: Make agc_weights work better across sample rate and n_fiber changes The agc_weights needed to be appropriately set up to do the right thing with different sample rates, and to allow different numbers of IHCs per channel. The AGC feedback is still proportional to the number of fibers, healthy or not, so the agc_weights parameters might need adjustment if the healthy_n_fibers is changed. The param n_fibers was renamed healthy_n_fibers, as that's what applied at design time, and as the n_fibers in the coeffs may change with model training (in the JAX version at least). -- 6cbf240 by Richard Lyon <[email protected]>: Delete .DS_Store COPYBARA_INTEGRATE_REVIEW=#18 from google:carfac-v3-start 6cbf240 PiperOrigin-RevId: 681626192
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.
Adding new SYN stage (synape model) for a v3 CARFAC.