MSE regression #98
Merged
MSE regression #98
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
…the per timestep errors that drive lambda_V. Added code to fill the ringbuffer and code for the drive of lambda_V.
… and runs but not yet successful in doing a regression. Also fixed n error in spike_recorder where the event recording flag was missing. Fixed two occurrences of pull_var_from_device.
… the pattern_recognition example. Everythig is compiling and running but there is no learning; in particular, no detectable systematic changes across epochs - not even in terms of regularisation of the hidden layer.
…e's fix for this. Changed how the per_times_step_loss code block is activated for loss_mse.
…into mse_regression
neworderofjamie
commented
Jun 27, 2024
Comment on lines
98
to
107
| ''' | ||
| for (pop,pypop) in compiled_net.neuron_populations.items(): | ||
| print(pypop.model.get_sim_code()) | ||
| for var in pypop.model.get_vars(): | ||
| print(var.name) | ||
| for para in pypop.model.get_params(): | ||
| print(para.name) | ||
| for dp in pypop.model.get_derived_params(): | ||
| print(dp.name) | ||
| ''' |
There was a problem hiding this comment.
Suggested change
| ''' | |
| for (pop,pypop) in compiled_net.neuron_populations.items(): | |
| print(pypop.model.get_sim_code()) | |
| for var in pypop.model.get_vars(): | |
| print(var.name) | |
| for para in pypop.model.get_params(): | |
| print(para.name) | |
| for dp in pypop.model.get_derived_params(): | |
| print(dp.name) | |
| ''' |
Comment on lines
120
to
122
| VarRecorder(output, genn_var="LambdaV", key="output_lambdav"), | ||
| VarRecorder(output, genn_var="LambdaI", key="output_lambdai"), | ||
| SynVarRecorder(h2o, genn_var="dw", key="h2o_dw"), |
There was a problem hiding this comment.
Not sure if we want to record all internal state in the example
… the eventprop pattern_recognition example.
…not be recorded correctly in batched simulations
* Added in corner case fixed from ``VarRecorder`` * Added ``get_underlying_conn`` helper to handle getting connectivity from ``Layer`` objects * Added source and target neuron masks for recording
Those all look good. Co-authored-by: neworderofjamie <[email protected]>
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #98 +/- ##
==========================================
+ Coverage 65.92% 66.08% +0.15%
==========================================
Files 100 101 +1
Lines 4024 4125 +101
==========================================
+ Hits 2653 2726 +73
- Misses 1371 1399 +28 ☔ View full report in Codecov by Sentry. |
tnowotny
approved these changes
Jun 27, 2024
of V versus target.
…into mse_regression
…ndoes the damage.
neworderofjamie
commented
Jul 1, 2024
Co-authored-by: neworderofjamie <[email protected]>
tnowotny
approved these changes
Jul 8, 2024
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.
This PR extends the existing mechanism in the EventProp compiler used to support 'per-timestep' cross-entropy loss to also supports mean-squared error loss for regression tasks. Using this, the pattern recognition task (previously used as an e-prop example) has been implemented as an EventProp example (encountering #101 and #102 in the process)
Also, this PR fixes a number of corner cases involving recording of non-batched variables in batched simulations and implements
ConnVarRecorderfor recording fromConnectionobjects.Fixes #52