London born, and NYC based Machine Learning Engineer at @meta. Particle physicist in a former life working on the @novaexperiment. He/him.
New York City
Joined November 2011
- Tweets409
- Following737
- Followers218
- Likes8.2K
Alexander retweeted
We're happy to share that @TmlrOrg recently accepted our paper on "Efficient CDF Approximations for Normalizing Flows" by Chandramouli Sastry, Andreas Lehrmann, @marcusabrubaker and @Alex_Radovic. You can read it here: bit.ly/3TxaTcL.
#CDF #normalizingflows #ML
Alexander retweeted
Efficient CDF Approximations for Normalizing Flows
Chandramouli Shama Sastry, Andreas Lehrmann, Marcus A Brubaker, Alexander Radovic
openreview.net/forum?id=Lnjc…
Alexander retweeted
I'm thrilled that our Machine Learning for Physical Sciences #NeurIPS2020 workshop proposal was accepted! Last year we had a great turnout and fantastic talks - check out the #ML4PS hashtag. New to team: @adjiboussodieng @iamstarnord @glouppe @zdeborova ml4physicalsciences.github.i…
ALT photos of website, organizers, and steering committee members
Alexander retweeted
OK, so, I've seen a lot of the "NASA detects parallel universe" stuff going around
I've spent the last ~2 years of grad school investigating these detections, and wanted to provide a bit of background and clarification
(1/n)
Alexander retweeted
In our preprint “The large learning rate phase of deep learning: the catapult mechanism" arxiv.org/abs/2003.02218, we show that the choice of learning rate (LR) in (S)GD separates deep neural net dynamics into two sharply distinct types (or "phases", in the physics sense). (1/n)
Alexander retweeted
A nice article by @Alex_Danco that explains how academic research really works, how it became the way it is, and how Twitter might play an important role in disrupting it.
Good read for non-academics to understand incentive structures of the institution.
danco.substack.com/p/can-twi…
Alexander retweeted
Our review, focused on some recent efforts at this interface and hopefully accessible to a non-ML science audience.
1/ Our new paper in @AnnualReviews of Condensed Matter Physics on “Statistical Mechanics of #DeepLearning” with awesome collaborators @Stanford and @GoogleAI : @yasamanbb @kadmonj Jeff Pennington @sschoenholz @jaschasd
web: annualreviews.org/doi/abs/10…
free: annualreviews.org/eprint/B3F…
Alexander retweeted
Did you know that other than the REINFORCE estimator and reparametrization trick, there is another Monte Carlo gradient estimator called the "measure-valued gradient estimator"? Check out my new blog post: bochang.me/blog/posts/measur…
Alexander retweeted
IMPORTANT: Students at UCSC are on strike. They're campaigning for a small (and urgently needed) Cost of Living Adjustment, or COLA. UCSC has responded with armed police and threats of deportation and firings.
Follow: @payusmoreucsc
Donate: tinyurl.com/paythemmoreucsc
Alexander retweeted
Happy to share our work on using generative models as surrogates of non-differentiable simulators to enable (approximate) gradient based optimization. arxiv.org/abs/2002.04632
Alexander retweeted
ML Fairness Gym is a set of components for building simple simulations that explore the potential long-run impacts of deploying machine learning-based decision systems in social environments, based on OpenAI Gym interface.
github.com/google/ml-fairnes…
Alexander retweeted
The Diversity-Innovation Paradox in Science: Why does greater diversity in research teams increase innovation but not directly reward minority scholars? New from @BasHofstra, @viveksck, Sebastian Munoz-Najar Galvez, Bryan He, @jurafsky, and Dan McFarland. arxiv.org/abs/1909.02063
Alexander retweeted
To make my research more reproducible, extensible and comparable to that of others & out of need to homogenize the language we use to express nn pruning methods, I contributed `nn.utils.prune` to #PyTorch 1.4 (see highlights github.com/pytorch/pytorch/r…)
Try it out & build on it! 🔥
v1.4: customizable mobile builds, Distributed Model Parallelism via experimental RPC API, Java Bindings, Chaining LRSchedulers
Summary: pytorch.org/blog/pytorch-1-d…
Release Notes: github.com/pytorch/pytorch/r…
Last release for Python 2 (bye bye!)
Alexander retweeted
Congrats to @elasri_layla for making the 2019 list of top women advancing #AI. Layla is leaving a mark in her field and on our teams at Borealis AI! bit.ly/36XuqJZ
Alexander retweeted
Right where we belong! Our researcher @Alex_Radovic on an adventure to take the Borealis AI gear home this season #NorthernLights #AuroraBorealis
Alexander retweeted
[1/2] Using WaveNet technology to reunite speech-impaired users with their original voices: in collaboration with @GoogleAI, we've applied new iterations including WaveRNN to help ALS advocate @TShawsTruth recreate & hear his original speaking voice
Blog: deepmind.com/blog/article/Us…
Alexander retweeted
Ready to kick off a great week at @NeurIPSConf. We’re currently hiring for a number of #ML positions. Come by our booth to find out more and meet our researchers. bit.ly/38lapPi #NeurIPS2019
Alexander retweeted
I tried AdverTorch from @BorealisAI for my class project on adversarial robustness of gradient penalized classifiers (arxiv.org/abs/1910.06922) and its very easy to use. You can plug-and-play and test accuracy at different levels of adversarial examples.
Great news! Our #AdverTorch toolbox has been added to the #PyTorch Ecosystem. Learn how to use it for attack-and-defence strategies here:
borealisai.com/en/blog/adver…
pytorch.org/ecosystem #AINorth
