Olivier Teytaud

You can improve GANs by evolutionary methods. See our cats at m.facebook.com/groups/nevergr… #eg3d #gan

@Slatefr change.org/p/pour-que-les…

@BFMTV change.org/p/pour-que-les…

@PapNdiaye J'aurais aimé pouvoir compter sur vous & que vous puissiez compter sur moi. Mais aujourd'hui, jour de pré-rentrée, je suis à la maison, en disponibilité. Deux départements m'accordent un INEAT mais le 92 dans lequel j'ai enseigné pendant 18 ans refuse de me donner un EXEAT.

Pour que les enseignants en disponibilité exercent dès la rentrée! - Signez la pétition ! chng.it/92JC8fHS via @ChangeFrance

@acversailles D. Fis, vous parlez de handicap ? "Tout enfant[...] présentant [...] un trouble invalidant de la santé est inscrit dans l'école [....] le plus proche de son domicile" Pourquoi pas pour notre enfant que vous voulez envoyer à 32 mins ? @acversailles #cned @jmblanquer

#Déconfinement Dominique Fis, directrice académique du 92 et Dominique Terrien, directeur académique adjoint, ont visité l’EREA Toulouse-Lautrec, à Vaucresson. Ils ont pu échanger avec des élèves de CP et CM2, en situation de handicap, ayant repris la classe depuis mardi.

@libe @SandrineMorch Notre enfant asthmatique sévère envoyé au collège à 32 min de marche de l'autre côté d'une autoroute + 2 nationales alors qu'il y en a un à 6 min sans gros axe. @SandrineMorch Merci DASEN92 @acversailles Au CNED pour sa santé.

La députée LREM @SandrineMorch présente son rapport pour identifier les freins à la scolarisation des enfants évoluant loin de l’école. Parmi ses préconisations : l’augmentation du nombre de médiateurs scolaires. bit.ly/3u3Kdps

@kairosdojo @rhyolight Nevergrad is certainly not a monster of computation, it's one of the computationally cheapest things going on these days in machine learning. It reaches state of the art performance on OpenAi Gym docs.google.com/document/d/1no… with a very small computational cost (no GPU needed).

@rhyolight Ohhh yeahh. I urge you then to give a look at those (basically bruteforce-Monte Carlo simulation) monsters of computation going under the names: "nevergrad", "neuroevolution", etc. . What upset me is that they still put the prefix "neuro-" in front of these words, as if a rhesus

The more I learn about the chain rule of composed derivatives and how backprop uses it to enable #DeepLearning, the more amazed I am at the incomprehensibly massive amount of brute force computation going on in most DL systems.

@spokutta @cian_neuro Now Nevergrad includes support for state of the art performance on OpenAi Gym docs.google.com/document/d/1no…

@cian_neuro You can also try scipy's basinhopping docs.scipy.org/doc/scipy-0.18… or gpyopt sheffieldml.github.io/GPyOpt/. Depending on how much time you have and function structure they work great. Nevergrad sounds also interesting - it is on my list to benchmark and write a post about it soon.

@FreakX19 @rasmusbergpalm @FacebookAI Now including support for state of the art performance on OpenAi Gym docs.google.com/document/d/1no…

I'm super pleased to release EvoStrat github.com/rasmusbergpalm… It's a library that makes Evolutionary Strategies (ES) simple to use. I'm really happy how cleanly it separates the environment, the reinforcement learning agent, the population distribution and the optimization.

@clagiuliano @GianniBarlacchi @GiacomoZara1 @AndreiCComan @BSearcher @Turchi_Marco @m_guerini @stefano_menini @Giuseppe_Jurman Now including support for state of the art performance on OpenAi Gym docs.google.com/document/d/1no…

Just found #Nevergrad: An #OpenSource tool for derivative-free optimization code.fb.com/ai-research/ne… #Phyton #AI #MachineLearning #DeepLearning @GianniBarlacchi @GiacomoZara1 @AndreiCComan @BSearcher @mdigangiPA @Turchi_Marco @m_guerini @stefano_menini @Giuseppe_Jurman

@Analyticsindiam Now including support for state of the art performance on OpenAi Gym docs.google.com/document/d/1no…

The Nevergrad platform enables AI researchers, machine learning scientists, and enthusiasts whose work involves derivative-free optimization to implement state-of-the-art algorithms.
 owl.li/bZvL30nVhiO
 
 #AI #ML #MachineLearning #Algorithms

@loretoparisi @FacebookAI Now including support for state of the art performance on OpenAi Gym docs.google.com/document/d/1no…

Optimizing #MachineLearning models training with the new @facebookai #Nevergrad github.com/facebookresear…

@ghb_hq Now including support for state of the art performance on OpenAi Gym docs.google.com/document/d/1no…

'nevergrad' by facebookresearch : A Python toolbox for perf... github.com/facebookresear… #dev #programming #tech #python

@iamchuckrussell Now including support for state of the art performance on OpenAi Gym docs.google.com/document/d/1no…

@iamchuckrussell Now including support for state of the art performance on OpenAi Gym docs.google.com/document/d/1no…

@skepteis Now including support for state of the art performance on OpenAi Gym docs.google.com/document/d/1no…

Facebook deserves a little bit of good press once in a while... 😉 Nevergrad: A Python library for performing derivative-free ML optimization - code.fb.com/ai-research/ne…

@DirkSchaar Now including support for state of the art performance on OpenAi Gym docs.google.com/document/d/1no…

NeverGrad: gradient free optimization library. Now open source. code.fb.com/ai-research/ne…

@JosephAEarly Now including support for state of the art performance on OpenAi Gym docs.google.com/document/d/1no…

@tunguz @Analyticsindiam Now including support for state of the art performance on OpenAi Gym docs.google.com/document/d/1no…

FB’s New Python Library Nevergrad Provides A Collection Of Algorithms That Don’t Require Gradient Computation analyticsindiamag.com/fbs-new-python… via @Analyticsindiam