-
Notifications
You must be signed in to change notification settings - Fork 18
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Neural Network submission to the challenge including 3x2 and FoM or SNR #9
Open
EiffL
wants to merge
43
commits into
LSSTDESC:master
Choose a base branch
from
EiffL:master
base: master
Could not load branches
Branch not found: {{ refName }}
Loading
Could not load tags
Nothing to show
Loading
Are you sure you want to change the base?
Some commits from the old base branch may be removed from the timeline,
and old review comments may become outdated.
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
Here are my results with the new Buzzard dataset, only on First the numbers. This is when optimizing for the
and this is when optimizing for the
And here are what these bins look like: |
Open
For reference, here are my final Buzzard results:
These list results for training with the FoM DETF or 3x2 SNR as the loss function |
And same thing for DC2:
|
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Outdated suggestions cannot be applied.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 presents a complete solution to the current challenge, including trained models, plots, and result metrics. It's an update of #4 for the new form of the challenge.
I haven't done any optimization of the model yet, and there is a high probability that I still have bugs somewhere, but wanted to share some of the results.
Here are the caveats:
riz
so farWhen optimized on the total 3x2 SNR, the neural network generally tries to build disjoint bins:

When optimizing on the FoM, it doesn't care that much and seems to like weird solutions where one large bin has contributions at both low and high redshift:

plots can be found in the
plots
folder, and results of the metric in the example folder. Note that I have noticed the FoM values out of cosmosis to be quite finicky so I don't trust them too much.