myelin

Machine Learning on Kubernetes

[Pau Rodríguez @prlz77]

#ICLR2020 submissions on graph neural networks, NLP and robustness have the greatest growth. @iclr_conf @openreviewnet

[Gabriel Peyré @gabrielpeyre]

A paper I enjoyed reading recently: Inverse problems in spaces of measures, Bredies, Pikkarainen. Introduces a Frank Wolfe algorithm with non convex improvement to solve infinite dimensional optimization over measures. A must read. https://t.co/SgDssJFb1N

[Timothée Lesort @TLesort]

I just created a pytorch compatible data loader specifically for Continual Learning on github I hope it might be useful for someone @ContinualAI
https://t.co/g2mromxFKH
#ContinualLearning #DeepLearning #lifelonglearning #OpenSource

[Gabriel Peyré @gabrielpeyre]

A paper I enjoyed reading recently: Breaking the Curse of Dimensionality with Convex Neural Networks, @BachFrancis (fig by @LenaicCsl ). Frames training of sparse 1 hidden layer MLP as optimization over measures, very inspiring. https://t.co/dnPZJvuhgF

[Gilles Louppe @glouppe]

New paper by my student @WehenkelAntoine: Unconstrained Monotonic Neural Networks! https://t.co/3bWLj8ck0l In this work, we show how to make use of free-form neural networks to model arbitrary monotonic functions and derive a new autoregressive flow.

[Kai Irandoust @k_irandoust]

OpenFaaS multi-stage with Istio on Kubernetes - by @AmirhosseinMov5
#OpenFaas #Serverless #Kubernetes #CertManager #Istio @openfaas
https://t.co/NbYzdhMizk

[Jigar Doshi @jigarkdoshi]

@fhuszar This @SimonsInstitute 'Boot Camp' has some really insightful talks/surveys about generalization, including the effect of batch_size on generalization https://t.co/k8LRfq8PfK

[Janos Matyas @matyix_]

Showtime - running an @IstioMesh across 5 cloud providers and 6 #Kubernetes distributions . Why? Because we can :) https://t.co/RzSGY2Ylvy

[Artëm Sobolev @art_sobolev]

Second post in the Mutual Information series: this time I address the notorious Formal Limitations paper and show how it manifests itself in different bounds.

https://t.co/KyR0Emhf3X

[Artëm Sobolev @art_sobolev]

Recently, I've been thinking about the Mutual Information and ways to estimate it. I'd like to share some of my thoughts and start with suggesting some novel / overlooked ways to estimate the MI.

https://t.co/7d62MT5GQB

[Bryan Catanzaro @ctnzr]

MLPerf 0.6 is out! Training speed on a single DGX-2 improved up to 1.8X over 7 months ago, and scaling efficiency to many nodes is markedly better.

The V100 GPU, 2 years after intro, is still the fastest and most general processor for deep learning.

https://t.co/mEDGgBGiIJ

[Jeremy Howard @jeremyphoward]

An amazing new project from @bearpelican was just released: https://t.co/DBov6sZTVS . A beautiful design; you can auto-generate a melody from chords, chords from a melody, and more.

It's technically brilliant, combining BERT, seq2seq, and Transformer XL
https://t.co/jF3mO5aXiu

[Bryan Catanzaro @ctnzr]

Here’s how we trained an 8.3B parameter GPT-2. We alternate row- and column- partitioning in the Transformer in order to remove synchronization and use hybrid model/data parallelism. 15 PFlops sustained on 512 GPUs. Details and code: https://t.co/7eXA6r15yX

[Google DeepMind @GoogleDeepMind]

We are excited to release Behaviour Suite for Reinforcement Learning, or ‘bsuite’ – a collection of carefully-designed experiments that investigate core capabilities of RL agents

GitHub: https://t.co/kGkRtFhezR
Paper: https://t.co/UJcALddJpc

[antonio vergari ⚔️ @tetraduzione]

@fhuszar a good ol' read in that regard from @jessejdavis1

https://t.co/TzkJOx1J20

they are very correlated but PR-AuC provides more information when the dataset is very unbalanced

[Gabriel Peyré @gabrielpeyre]

Oldies but Goldies: D. Donoho, Compressed sensing, 2004. Introduced (jointly with Candes, Romberg, Tao) the compressed sensing idea. https://t.co/Vz1PCMEifk

[Gabriel Peyré @gabrielpeyre]

Gradient descent on particles’ positions (Langrangian) is equivalent to an advection linear PDE on the density of particles (Eulerian). https://t.co/lHYMjNfyZt