tensorflow / tensorflow/probability

Sample different weights for examples within a batch

Open
#153 9 comments 2 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

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

Description

If I'm not mistaking currently bayesian Conv2d samples single kernel for all examples within a batch. In principle sampling different kernels for each of the examples could make MC estimates better. This requires batched convolution where the kernel has batch dimention too. Possible approach is here:
https://stackoverflow.com/questions/42068999/tensorflow-convolutions-with-different-filter-for-each-sample-in-the-mini-batch
Does it worth to implement such feature?

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 locating the current Bayesian Conv2d implementation and understanding how it samples one kernel for a batch. Review the linked TensorFlow convolution approach for per-example filters. Done means agreeing on the feature scope and implementing per-example kernel sampling with coverage for the resulting MC estimates; no file or test is named here.

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
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.