Skip to content
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.

Sign up for GitHub

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

Jump to bottom

Investigate Keras Model Compression before SQL generation #5

Open
antoinecarme opened this issue Jul 7, 2018 · 4 comments
Open

Investigate Keras Model Compression before SQL generation #5

antoinecarme opened this issue Jul 7, 2018 · 4 comments
Assignees
@antoinecarme

Comments

@antoinecarme
Copy link
Member

antoinecarme commented Jul 7, 2018 •
edited
Loading

This is probably a way to avoid some unnecessarily large models and generate an (almost) equivalent SQL code.
@antoinecarme antoinecarme changed the title Inmvestuagte Compress the Keras Model before SQL generation Investigate Model Compression before SQL generation Jul 7, 2018
@antoinecarme antoinecarme changed the title Investigate Model Compression before SQL generation Investigate Keras Model Compression before SQL generation Jul 7, 2018
@antoinecarme
Copy link
Member Author

There are at least 4 methods for compressing keras models in

image

https://arxiv.org/pdf/1710.09282.pdf

@antoinecarme
Copy link
Member Author

image

@antoinecarme antoinecarme mentioned this issue Jul 7, 2018
Open
@antoinecarme
Copy link
Member Author

Would be nice to test this :

Model Compression Based on Geoffery Hinton's Logit Regression Method in Keras applied to MNIST 16x compression over 0.95 percent accuracy

https://github.com/Irtza/Keras_model_compression

@antoinecarme
Copy link
Member Author

Factorization method :

low-rank approximation
1. with SVD (matrix)
2. with Tucker (tensor)

https://github.com/DwangoMediaVillage/keras_compressor

@antoinecarme antoinecarme self-assigned this Jul 18, 2019
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@antoinecarme antoinecarme

Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@antoinecarme