# Differentiating random walk probability w.t.r. rate of jump

**URL:** https://discourse.julialang.org/t/differentiating-random-walk-probability-w-t-r-rate-of-jump/107116
**Category:** Probabilistic Programming
**Tags:** autodiff
**Created:** [December 4, 2023, 2:46pm UTC](https://discourse.julialang.org/t/differentiating-random-walk-probability-w-t-r-rate-of-jump/107116 "2023-12-04T14:46:10Z")
**Posts on this page:** 1
**Showing post:** 12

<div class="post-metadata">

### Author: ![mlanghinrichs](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/mlanghinrichs/32/50371_2.png) [@mlanghinrichs](https://discourse.julialang.org/u/mlanghinrichs)
#### Post date: [December 14, 2023, 6:56pm UTC](https://discourse.julialang.org/t/differentiating-random-walk-probability-w-t-r-rate-of-jump/107116/12 "2023-12-14T18:56:47Z")

</div>

Super, that’s amazing!

I had to go into ABC-SMC algorithms myself this year, as we wanted to keep track of the “particle weights” as a model evidence estimate for model selection (or normalising constant, or marginal likelihood). Seems to work quite well for problems I encountered so far ([ABCdeZ.jl](https://discourse.julialang.org/t/ann-abcdez-jl-bayesian-inference-model-evidences/99655)).

From this point of view it would be really fantastic to have gradient-based ABC-SMC that can still keep track of the correct (readjusted maybe) particle weights, to not loose the model evidence estimate. Feels like it should be possible, but don’t know for sure. Happy to join or distribute forces if anything would help.

---

_[View the full topic](https://discourse.julialang.org/t/differentiating-random-walk-probability-w-t-r-rate-of-jump/107116)._