tensorflow / tensorflow/probability

how do you make variable-sized ("ragged") output distributions with tfp.layers?

Open
#523 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Jupyter Notebook
Stars
4.4k
Forks
1.1k
PR merge metrics
No merged PRs in 30d

Description

We'd like to model a molecular problem with variable number of atoms.

How do you make a variable-sized / ragged distribution shape with tfp layers?

Time series doesnt make sense here. We'd like a distribution output so we can use the nice built in features of tfp distribution like entropy, etc. Tried Convolution1DFlipout but this retuens a tensor, not a distribution. Do we need to set batch_size=n_atoms?

Sorry if this is (another) noob question
Bionicles

Contributor guide

Open the contributing guide

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

Start by reading the tfp.layers distribution-producing layers and the Convolution1DFlipout API mentioned in the issue. Check how variable-sized or ragged inputs and distribution outputs are currently handled, then determine whether the requested behavior needs documentation or a new feature. Done means establishing a supported approach for variable numbers of atoms or clearly documenting that it is not supported.

Written by the indexing model from the issue text.

Assessment

Tech stack
tensorflow
Domain
machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.