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.

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

StoDIG submission #5 #222

Merged
merged 1 commit into from
Feb 1, 2017
Merged

StoDIG submission #5 #222

merged 1 commit into from
Feb 1, 2017

Conversation

da-wad
Copy link
Contributor

@da-wad da-wad commented Jan 31, 2017

Our new Facies Probability plots have helped us in understanding what's happening, so the 5th (and final) submission from StoDIG has two models we'd like to enter, and we'd be fascinated to see how they score... please find two notebooks:

  • Firstly an increment to our CNN which includes distance to next Formation as a feature
  • Secondly a new innovation, using an LSTM network along with the CNN

The work in the last weeks has been a really great opportunity to have a hands-on experience with Neural Networks, so thanks to the organisers from StoDIG!

@kwinkunks kwinkunks merged commit 0c8d9f2 into seg:master Feb 1, 2017
@kwinkunks
Copy link
Member

Hello! ...5_CLDNN scores 0.549, while ...5_CNN scores 0.541.

Thank you for the appreciation... I'm glad you had fun, and even better that you were learning. That's a big win!

The combined LSTM / CNN is really interesting. I'm a big fan of LSTMs, but have yet to get them to actually help on a stratigraphic problem. I have a hard time believing that this sort of model wouldn't beat the 'shallow' methods on a very large dataset. It might be hard to find an open dataset that size, but I'm sure you have something at work to test it on!

Cheers, Matt

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants