Papers of the day   All papers

Monte Carlo Gradient Estimation in Machine Learning

Comments

Shakir Mohamed: Exited to share our new paper: 'Monte Carlo Gradient Estimation in Machine Learning', with @elaClaudia @mfigurnov @AndriyMnih. It reviews of all the things we know about computing gradients of probabilistic functions. https://arxiv.org/abs/1906.10652 🐾Thread👇🏾 https://t.co/2eTPsFO7mZ

11 replies, 1073 likes


Shimon Whiteson: I highly recommend this excellent survey paper on gradient estimation.

1 replies, 212 likes


Been Kim: Gradients are important for interpretability! This is a very accessible, high quality review/tutorial on how to estimate them, with code!

1 replies, 206 likes


Statistics Papers: Monte Carlo Gradient Estimation in Machine Learning. http://arxiv.org/abs/1906.10652

0 replies, 124 likes


Shakir Mohamed: Following the outpouring of support for #JMLR🤩, excited to see our paper on Monte Carlo Gradient Estimation on the JMLR site 🥳 http://jmlr.org/papers/v21/19-346.html Still lots to learn in this space and grateful for all the ongoing recommendations and feedback from everyone. 🐝

1 replies, 109 likes


Joey Bose: Wow this is perhaps the best review of stochastic gradient estimators I've ever seen. Definitely, a must read for all people in doing VI, RL, and other cool ML stuff.

0 replies, 80 likes


Mihaela Rosca: The code reproducing the experiments is this paper is now available at: https://github.com/deepmind/mc_gradients

0 replies, 77 likes


Mathias Niepert: Monte-Carlo gradient estimation article https://arxiv.org/abs/1906.10652 has been super enlightening and fun to read. After studying it we improved section on gradient estimation of our ICML paper on jointly learning graph structures and GNNs. @shakir_za @MihaelaCRosca @AndriyMnih

6 replies, 62 likes


Mihaela Rosca: I really enjoyed working on this review paper with my amazing co workers! So much to learn!

1 replies, 59 likes


Adji Bousso Dieng: @yoavgo Some of my favorite review papers (recent): MC gradient estimation in ML by @shakir_za et al. https://arxiv.org/abs/1906.10652 Normalizing flows by @gpapamak et al. https://arxiv.org/abs/1912.02762 Variational inference (pre-amortization) by Blei et al. (@blei_lab) https://arxiv.org/abs/1601.00670

2 replies, 54 likes


Shakir Mohamed: After a short delay, the code in a notebook to reproduce the graphs in section 3 of our paper (https://arxiv.org/abs/1906.10652) is online. More to be come soon. See thread above👆🏾. https://github.com/deepmind/mc_gradients 👩🏾‍💻

0 replies, 52 likes


Pietro Lesci: As the ML field is moving VERY quickly, review papers are a true need and great opportunity to learn and get new insights. Recommended!

0 replies, 34 likes


Debasish (দেবাশিস্) Ghosh 🇮🇳: a wonderful survey of stochastic gradient estimators by @shakir_za and team https://arxiv.org/abs/1906.10652. Very neat derivations using simple probabilistic tricks and clear explanations. Summary in this twitter thread https://twitter.com/shakir_za/status/1143802522299244545

0 replies, 34 likes


ML Review: Monte Carlo Gradient Estimation in Machine Learning [59pp] By @shakir_za @elaClaudia @mfigurnov Survey of methods for Monte Carlo gradient estimation: the pathwise, score function, and measure-valued gradient estimators. https://arxiv.org/abs/1906.10652

0 replies, 29 likes


Shakir Mohamed: #WritingReflections✍️🏾: The paper I looked forward to most was our paper on Monte Carlo Gradient Estimation. Took 1 yr to write, with the aim to uplift our readers; to summarise 50yrs of broad research. Still prob the best paper i've written 🐾https://twitter.com/shakir_za/status/1143802522299244545?s=20

1 replies, 24 likes


Bo Chang: It is based on the great survey paper "Monte Carlo Gradient Estimation in Machine Learning" by @shakir_za, @MihaelaCRosca, @mfigurnov, and @AndriyMnih: https://arxiv.org/abs/1906.10652

0 replies, 23 likes


Sourav Mishra: “One paper to review them all” Very exhaustive survey of Gradient Estimation!

0 replies, 13 likes


Jiaxin Shi: Cool survey of Monte Carlo gradient estimation. Also thanks for including our log density gradient estimators! @shakir_za

0 replies, 12 likes


Jacky Liang: @zacharylipton @jachiam0 found this to be a very useful review/overview - "Monte Carlo Gradient Estimation in Machine Learning" by @shakir_za https://arxiv.org/abs/1906.10652

0 replies, 11 likes


Emtiyaz Khan: I thank @shakir_za and colleagues to write this wonderful paper about gradient estimators. Difficulty of estimating these is one of the main reasons behind the difficulties in many fields such as Bayesian inference, RL etc.

0 replies, 6 likes


Matias Quiroz: This goes to the top of my to-read-list!

1 replies, 5 likes


Colin Carroll: Love when papers come with threads like this, especially the photo-tour of highlights!

0 replies, 4 likes


Thanos Papaoikonomou: "Monte Carlo Gradient Estimation in Machine Learning" https://arxiv.org/pdf/1906.10652.pdf by @shakir_za @elaClaudia @mfigurnov @AndriyMnih

0 replies, 4 likes


Thanos Papaoikonomou: @DeepMind That's great work! I love the fact that it includes stochastic gradient estimators based on https://arxiv.org/abs/1906.10652 . Thanks for releasing these awesome tools :)

0 replies, 4 likes


Rory Quinn: A survey of methods for Monte Carlo gradient estimation #machinelearning https://arxiv.org/abs/1906.10652?utm_campaign=Data_Elixir&utm_medium=email&utm_source=Data_Elixir_240

0 replies, 1 likes


Content

Found on Jun 26 2019 at https://arxiv.org/pdf/1906.10652.pdf

PDF content of a computer science paper: Monte Carlo Gradient Estimation in Machine Learning