Dimitris Korkinof

Machine Learning at Spotify - my views are my own

[Dimitris Korkinof @korkinof]

Being pushed by the reviewers of #AIRad to come up with something better than eyeballing for evaluating the distributional alignment of #GANs, we proposed a very similar plot based on image moments. It's not only a good visualisation of ultra-high dimensional distributions... 1/6

[KDnuggets @kdnuggets]

7 Must-Try #DataVisualization Libraries in #Python https://t.co/Oq0cjaK5Y0

[Jure Leskovec @jure]

📢 New research alert! 💡 We've developed PRODIGY: pretraining framework for in-context learning over graphs. Through a novel prompt graph representation and a family of in-context pretraining objectives, our model can adapt to novel tasks on unseen graphs 📈. Outperforming contrastive pretraining baselines by 18% and standard finetuning with limited data by 33% on average, PRODIGY proves its strength in citation networks and knowledge graphs. Read on: https://t.co/36HtroWybR #AIResearch #GraphLearning #PRODIGY #InContextLearning

Joint work with @qhwang3 @ren_hongyu PengChen GregorKrzmanc DanielZheng and @percyliang

[Alvaro Cintas @dr_cintas]

This is impressive.

META just released MusicGen, a Language Model designed for creating music.

Not just that, it produces high-quality music while being conditioned on text description or melodic features.

Best thing?

You can try it FREE now.

Here is a Demo of converting the famous Bach melody “Toccata and Fugue in D Minor” into an 80s driving pop song.

[Dimitris Korkinof @korkinof]

I’m happy to share that I've been promoted to Staff ML Engineer at Spotify. https://t.co/MyZZOSv62r

[Felix Petersen @FHKPetersen]

I am excited to announce that I have successfully defended my PhD and have published my PhD thesis on “Learning with Differentiable Algorithms”. 🎉

https://t.co/UkvS6KQ1uX

In the thesis, I explore how we can make discrete structures like algorithms differentiable. [1/13]

[Dimitris Korkinof @korkinof]

https://t.co/jwoFSCi0jX https://t.co/9Y4v8oy1BV

[Stanford NLP Group @stanfordnlp]

Like we said, the author list of the foundation models paper is completely reasonable…

[Mick Ryan, AM @WarintheFuture]

“Russia must suffer such a devastating defeat that it will be decades before another Russian leader thinks of attacking a peaceful neighbor.” An outstanding article from ⁦@MaxBoot⁩ on the war in #Ukraine, and the necessity of a Russian defeat. https://t.co/MWNZs0357y

[Dimitris Korkinof @korkinof]

https://t.co/TliNvTy8Rn

[Dimitris Korkinof @korkinof]

https://t.co/GHrHC36S6Q

[Kevin Patrick Murphy @sirbayes]

I am delighted to announce that a draft of my latest book, “Probabilistic Machine Learning: Advanced Topics”, is now available online at https://t.co/dSlKkwYpLr. It covers #DeepGenerativeModels, #BayesianInference, #Causality, #ReinforcementLearning, #DistributionShift, etc.

[David Pfau @pfau]

OK, this is one I’ve been waiting to share for a *long* time – the first ever demonstration of deep reinforcement learning on a nuclear fusion research device! https://t.co/HVfUCrMTlM

[Dimitris Korkinof @korkinof]

Oh, wait, actually the best Wordle opener is not “crane”… https://t.co/ifdb40Kb9p via @YouTube

[Kevin Patrick Murphy @sirbayes]

I am delighted to announce that my new book, “Probabilistic Machine Learning: An Introduction”, is finally available in print format! You can order it from https://t.co/fx92WBQvk3, or from Amazon. Also available at https://t.co/dSlKkwYpLr 1/4

[Dimitris Korkinof @korkinof]

The mathematically optimal Wordle strategy https://t.co/bBafWdOMxI via @YouTube

[Shane Gu @shaneguML]

Can pre-trained language models be used for offline RL? We look to answer this question in our new work and demonstrate SoTA-level performance on various offline RL benchmarks when adapting pre-trained LMs for RL 🤯

paper: https://t.co/gbG9Xy7grJ
code: https://t.co/iKYZsXrUPI 1/

[Ruiqi Gao @RuiqiGao]

Check our work "Learning Energy-based Model with Flow-based Backbone by Neural Transport MCMC"(w/ @erik_nijkamp, etc.): https://t.co/evg1ltsrHN.
We learn EBM by exp. tilting of flow, with mixing MCMC chains for 2000+ steps! Very excited that long run MCMC finally works for EBM.

[I Can't Believe It's Not Better! @ICBINBWorkshop]

It is official! We have launched the ICBINB initiative! (https://t.co/5y5KwdycX9) Let's crack open the research process together. Thank you @AaronSchein
@franciscuto
@_hylandSL
@in4dmatics
@wellingmax
@ta_broderick
and Bob Williamson to make it happen!!

[Hugo Larochelle @hugo_larochelle]

Today, @RaiaHadsell, @kchonyc and I are happy to announce the creation of a new journal: Transaction on Machine Learning Research (TMLR)

Learn more in our post: https://t.co/yviHASS4R1

[Mounia Lalmas @mounialalmas]

A great article about music recommendations at Spotify, including the research we are doing at @SpotifyResearch in this area.
https://t.co/jiCC090OLC

[Google DeepMind @GoogleDeepMind]

A new paper in @Nature details how machine learning was used to make significant new discoveries in pure mathematics by guiding the intuition of some of the world’s top mathematicians from @Sydney_Uni and @OxUniMaths: https://t.co/yWca5JOrHs

Paper: https://t.co/AvX78jY9zC 1/