Gaurav Arya

Programming languages, statistics, and gradients. CS PhD Student at Carnegie Mellon University

[Gaurav Arya @NotGauravArya]

Excited about this new working paper! Monte Carlo samplers are a wonderfully interesting class of discrete+continuous random processes, and taking their derivative can have a number of applications 🙂

[Stat.ML Papers @StatMLPapers]

Differentiating Metropolis-Hastings to Optimize Intractable Densities. (arXiv:2306.07961v1 [https://t.co/zjV5HgYw5a]) https://t.co/9aKMeb7C8H

[Gaurav Arya @NotGauravArya]

@MoritzSchauer @xuanalogue The code snippet in https://t.co/HKOQEKqC1E would be the closest thing for now🙂 but you've gotta wrap the strings branches etc. into a nice vector space -> vector space function:) For more flexibility, you'd want continuation-passing style like in ADEV https://t.co/nvRLiZuCMG!

[Gaurav Arya @NotGauravArya]

@AstroClary @l_II_llI not enough books though

[Gaurav Arya @NotGauravArya]

@cosmic_mar @MilesCranmer Makes sense. In that case, I guess you just have to hope that the subtypes of one of the two parent types can be finitely enumerated, and forward all the requisite methods onto each of them as @MilesCranmer noted. Sort of a "manually strapped on" trait dispatch 🙃

[Gaurav Arya @NotGauravArya]

@MilesCranmer @cosmic_mar But in the common case where we can't do that, we have to work with the (successful but sometimes flawed) default choice of dispatch algorithm. Another relevant link: https://t.co/w6bNDXUv6G

[Gaurav Arya @NotGauravArya]

@MilesCranmer @cosmic_mar meaning if one were to own a function, one has the power to change the dispatch paradigm (e.g. by asking users to define _f(args...) for their types, and then writing f(args...) to do arbitrary type-space manipulations to call the right _f implementation).

[Gaurav Arya @NotGauravArya]

@thatpinkney There's been some prior work (cited in our preprint) that's had some success doing only the former and not the latter, so it's an interesting space

[Gaurav Arya @NotGauravArya]

Excited about this new working paper! Monte Carlo samplers are a wonderfully interesting class of discrete+continuous random processes, and taking their derivative can have a number of applications 🙂

[Stat.ML Papers @StatMLPapers]

Differentiating Metropolis-Hastings to Optimize Intractable Densities. (arXiv:2306.07961v1 [https://t.co/zjV5HgYw5a]) https://t.co/9aKMeb7C8H

[Gaurav Arya @NotGauravArya]

@thatpinkney It's a natural next step! That involves differentiating through both HMC's continuous dynamics and the MH accept/reject. Some discussion here: https://t.co/gsvhwjf0OG

[xuan (ɕɥɛn / sh-yen) @xuanalogue]

@ChrisRackauckas @alexanderklew @_k_a_c_h_ @NotGauravArya Yup! I tend to think of HMC as "a fancy version of MH with a custom proposal probabilistic program", so I imagine the main additional work is propagating gradients through the deterministic parts of that program (i.e. the leapfrog integrator)!

[Moritz Schauer @MoritzSchauer]

📣 Announcing a new working paper on 𝒟𝒾𝒻𝒻𝑒𝓇𝑒𝓃𝓉𝒾𝒶𝒷𝓁𝑒 Metropolis-Hastings https://t.co/Q1tMwHTsHl 🧵

[Stat.ML Papers @StatMLPapers]

Differentiating Metropolis-Hastings to Optimize Intractable Densities. (arXiv:2306.07961v1 [https://t.co/zjV5HgYw5a]) https://t.co/9aKMeb7C8H

[Gaurav Arya @NotGauravArya]

Check out @MoritzSchauer's announcement!

https://t.co/nKbBKpUBgV

[Moritz Schauer @MoritzSchauer]

📣 Announcing a new working paper on 𝒟𝒾𝒻𝒻𝑒𝓇𝑒𝓃𝓉𝒾𝒶𝒷𝓁𝑒 Metropolis-Hastings https://t.co/Q1tMwHTsHl 🧵

[Gaurav Arya @NotGauravArya]

Our 4-page paper presents some key ideas on how to get there.

In particular, we aim to form a *consistent* derivative estimator a la the usual Ergodic theorem.

[Gaurav Arya @NotGauravArya]

Had my first poster session today, at #NeurIPS2022🙂 Photo credits @_Frank_Schaefer

[Gaurav Arya @NotGauravArya]

@ayushchopra96 @ChrisRackauckas @MIT @MoritzSchauer Hey, I agree they're related, and we'd be happy to cite your work in our final revision! Our hope is to differentiate these models unbiasedly, but as its empirical success shows Gumbel-Softmax is no doubt a good approach that avoids the complications of discreteness.