tensorflow / tensorflow/probability
tfp.mcmc.sample_halton_sequence for float32 turns bad at index 16.8 million
Open
Nobody has claimed this yet.
- Dominant language
- Jupyter Notebook
- Stars
- 4.4k
- Forks
- 1.1k
- PR merge metrics
- No merged PRs in 30d
Description
It appears that there is some kind of bit overflow at halton index 2**24 (16.8 million) when using tf.float32 as output:
The same pattern does not happen for float64:
I would venture a guess that this has something to do with the bit-reversal of Halton sequence and the float mantissa size...
A similar but different change happens for the second dimension.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.