@diffprogramming profile picture
Differentiable Programming @diffprogramming

Differentiable Programming @NeurIPSConf

Joined July 2021
  • Tweets 57
  • Following 80
  • Followers 468
13 Photos and videos
Exciting-looking program by @createaiplans today! Starts shortly; tune in!
AIPLANS Workshop@createaiplans
14 Dec 2021
The day we've all been waiting for has finally arrived! Join us today from 7am-6pm EST for a workshop full of possibilities at the intersection of programming languages and neurosymbolism. Brought to you in part by @Mila_Quebec, @NeurIPSConf and ML/PL enthusiasts just like you!
4
Differentiable Programming retweeted
AIPLANS Workshop@createaiplans
14 Dec 2021
The day we've all been waiting for has finally arrived! Join us today from 7am-6pm EST for a workshop full of possibilities at the intersection of programming languages and neurosymbolism. Brought to you in part by @Mila_Quebec, @NeurIPSConf and ML/PL enthusiasts just like you!
4
10
Wow, the day flew by! We are about the start the last session in less than 10 min. Come join us for talks by @sokrypton, Harshitha Menon, Gordon Plotkin, and Bhupalee Kalita!
1
2
11
more replies
The penultimate talk for the day: Gordon Plotkin on complete axiomatization of forward differentiation.
1
1
2
And our final talk today: thank you, Bhupalee Kalita, for your paper and talk on differentiable programming for spin-polarized systems! Check out the paper here arxiv.org/abs/2110.14846

How Well Does Kohn-Sham Regularizer Work for Weakly Correlated Systems?

Kohn-Sham regularizer (KSR) is a differentiable machine learning approach to finding the exchange-correlation functional in Kohn-Sham density functional theory (DFT) that works for strongly...

arxiv.org
1
2
We just kicked off our next session! Excited to hear from @PatrickHeimbach, Karen Liu, Nacime Bouziani, and @alexanderklew!
1
1
3
more replies
Nacime Bouziani spoke about a foreign function interface for the Unified Form Language [UFL]! See the paper here: arxiv.org/pdf/2111.00945.pdf

2
And @alexanderklew wrapped up this session with a talk on denotational Semantics of AD for higher-order, recursive PPLs. Make sure to check out the paper: arxiv.org/abs/2111.15456

Towards Denotational Semantics of AD for Higher-Order, Recursive,...

Automatic differentiation (AD) aims to compute derivatives of user-defined functions, but in Turing-complete languages, this simple specification does not fully capture AD's behavior: AD sometimes...

arxiv.org
Our next session is in 10min! Tune in for talks by Frank Noe (@FrankNoeBerlin), Keno Fischer (@KenoFischer), Patrick Kidger (@PatrickKidger), and Deniz Bezgin!
1
1
12
more replies
@PatrickKidger spoke about Equinox: a library that brings more power to your model building in JAX! Check it out here github.com/patrick-kidger/eq…

GitHub - patrick-kidger/equinox: Elegant easy-to-use neural networks scientific computing in JAX....

Elegant easy-to-use neural networks scientific computing in JAX. https://docs.kidger.site/equinox/ - patrick-kidger/equinox

github.com
1
1
4
Deniz Bezgin wrapped up the session describing a fully-differentiable compressible high-order computational fluid dynamics solver. See the paper here arxiv.org/abs/2112.04979

A fully-differentiable compressible high-order computational fluid...

Fluid flows are omnipresent in nature and engineering disciplines. The reliable computation of fluids has been a long-lasting challenge due to nonlinear interactions over multiple spatio-temporal...

arxiv.org
We are about to start the poster session very soon! Come join us in Gather Town! eventhosts.gather.town/app/d…
1
3
We are about to kick off the meeting in 1 hour! Join us for an exciting first session featuring Adam Paszke (@apaszke), Yuan Zhou (@yuaanzhou), Uwe Naumann, and Antonio Stanziola!
1
4
16
more replies
Thank you, Uwe Naumann for your talk about Differentiable Scripting. Check out Uwe's system here: github.com/un110076/AD4SL

GitHub - un110076/AD4SL

Contribute to un110076/AD4SL development by creating an account on GitHub.

github.com
1
1
3
Antonio Stanziola wrapped up the first session with a talk about jaxdf: a JAX-based package defining a coding framework for writing differentiable PDE discretizations! Check it out: github.com/ucl-bug/jaxdf

GitHub - ucl-bug/jaxdf: A JAX-based research framework for writing differentiable numerical...

A JAX-based research framework for writing differentiable numerical simulators with arbitrary discretizations - ucl-bug/jaxdf

github.com
2
3
Load more