tensorflow / tensorflow/probability

KL-reweighting not done correctly in BNN code ?

Open
#719 3 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

Hi,

It seems that there is some discrepancy in how the KL-reweighting is done
during stochastic/batched training.

In section 3.4 of this paper: Weight Uncertainty in Neural Networks, the KL divergence is re-weighted using "number of batches", but in the BNN example code here, the KL divergence is re-weighted using size of batch

Does this seem like a bug ?

Thanks

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

Read section 3.4 of the linked Weight Uncertainty in Neural Networks paper, then inspect the Bayesian neural network example at the referenced line in tensorflow_probability/examples/bayesian_neural_network.py. Compare the KL reweighting factors used for stochastic or batched training and determine whether the example matches the paper; done means resolving whether the discrepancy is a bug.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, tensorflow
Domain
machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.