Arnie

ML at @google, prev at @flipboard,@onepeloton co-organizer of PyData/AI-in-Production meetups
Migrating to @[email protected]

[Arnie @abvankenobi]

A case for embedding in recommender systems (new blog post)! https://t.co/fCka58YKA6 #MachineLearning #recommendersystems

[Justin Basilico @JustinBasilico]

The #RecSys2023 industry track is open for submissions! Its a great way to share your latest insights in real-world recommendation systems with the community as a talk or poster. Extended abstracts are due June 1st: https://t.co/jFzmEIIqIp

[Arnie @abvankenobi]

Also moving to Mastodon - link in bio.

[Arnie @abvankenobi]

Awesome work @decisiontree7 and team :)

[Shoya Yoshida @decisiontree7]

My blog post about learnings from building out Context-Aware Recommender Systems is now up on Peloton Engineering Blog! Please check it out! :)
https://t.co/Nbhy1wIfun

[Arnie @abvankenobi]

Merlin models/systems are starting to look really cool =)

[Shoya Yoshida @decisiontree7]

I've been playing around with NVIDIA's Merlin Models library for the last couple days for the H&M Kaggle competition and here are my thoughts on the library! 🧵

[Arnie @abvankenobi]

Another great post by @eugeneyan! It's a great introduction for anyone interested in counterfactual evaluation of interactive systems

[Eugene Yan @eugeneyan]

Do you find it weird that we evaluate recommendation systems on static interaction logs? I did.

I discuss why it's weird (observational vs. interventional) and how counterfactual evaluation & inverse propensity scoring helps simulate A/B tests offline.

https://t.co/EmXy7PmkP7

[Arnie @abvankenobi]

So excited to see my wonderful teammates and friends @g3shandar and @decisiontree7 present at @NVIDIAGTC tomorrow!!! Please stop by! Expect an incredible talk on how we are using NVIDIA Merlin to modernise our recommender systems at @onepeloton :D

[Arnie @abvankenobi]

@dcreemer Nooo... what will it take to keep you motivated?

[Arnie @abvankenobi]

@dcreemer Wow, your variable names actually make sense. Can't say that about the garbage I came up with.

[Arnie @abvankenobi]

@dcreemer @yosoysami @louiedinh @djovan0 @ajyells @grippy Challenge accepted. Now I'd be doing advent of code on the two days I was planning to ... not.

[Arnie @abvankenobi]

@aftncanada Normally, I'd agree as I have generally hated the implementation of VAR in England - but I'd equally hate losing a cup final because of an offside. I don't think this was a millimetre/centimetre call, it looked offside before the lines were drawn.

[Arnie @abvankenobi]

I was a bit curious today - Google always has the answer!

[Sibin @sibinmohan]

I’m seeing a massive disconnect (more so than usual) between the two twitter worlds I usually inhabit — the US and India. It is important to bridge the two now for a very important reason: #CovidIndia

Please read the whole thread and amplify.

THREAD

1/17

[Even Oldridge @Even_Oldridge]

Excited to share my latest blog which focuses on why recommender systems aren't faster on GPU and how you can fix that. TLDR; RecSys models are different from NLP and CV and require specialized dataloaders to get data to the GPU efficiently. https://t.co/Luey8PNy0E

[Arnie @abvankenobi]

There are sooo many posters at Neurips, my fingers are tired from all the walking around. #NeurIPS2020 #NeurIPS

[Arnie @abvankenobi]

(Not) Shocked that none of our algorithms thought this was an important story. Why is this considered not important enough to be Trending on twitter or be one of the major headlines on any North American media?

[The Associated Press @AP]

Afghanistan’s Interior Ministry says an hours-long siege of the Kabul University is over, with at least 19 dead and 22 wounded. The ministry’s spokesman said there were three attackers involved, and that all of them were killed. https://t.co/FcxzIndgF9

[Arnie @abvankenobi]

One of my highlights of this year's REVEAL at #recsys2020 was this amazing work presented by @usait0 - Open Bandit Pipeline - A library for bandit algorithms and off-policy evaluation as well some relevant data with propensity scores! https://t.co/AgqVKyIaFh