RosettaCommons / RosettaCommons/RFdiffusion

How to describe the loss at different timesteps?

Open
#163 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
3.1k
Forks
644
PR merge metrics
No merged PRs in 30d

Description

@joewatchwell @lyskov
This is really excellent work. I have some questions regarding this work.

When training the model, how do you appropriately describe the loss for protein trajectories? Is the loss treated equally for each timestep t? When t approaches T, the structural loss should be relatively large, while it should be relatively small when t approaches 1.

In fact, when t=T or t=1, real protein structures are used for supervision. This seems unreasonable or difficult, as it was impossible to obtain the same protein structure as the label when t=T.

Looking forward to your reply!

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

No file, test, or entry point is named. Start by locating the training loss implementation and timestep handling, then determine whether the issue can be answered from the existing behavior; done would require a documented explanation of loss weighting and supervision at t=1 and t=T.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
bioinformatics, machine-learning
Issue type
Documentation
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
15/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.