tensorflow / tensorflow/probability

Custom gradients in HMC?

Open
#904 17 comments 1 reaction 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

When passing a target_log_prob_fn that's not built from TF primitives (and hence doesn't allow for automatic differentiation) to the HamiltonianMonteCarlo kernel, is there a way to also pass a custom gradient function? Of course, you lose all the performance benefits of autodif, but this would significantly increase the potential areas of application for TFP's HMC.

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 reviewing the HamiltonianMonteCarlo kernel API and how target_log_prob_fn is currently differentiated. The issue does not name implementation files or tests, so first determine where gradient handling is defined and how a custom gradient would be represented. Done means a supported custom-gradient path with coverage for non-TensorFlow target functions.

Written by the indexing model from the issue text.

Assessment

Tech stack
machine-learning
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.