Aurélien Geron

Author of the book Hands-On #MachineLearning with #ScikitLearn, #Keras and #TensorFlow. Former PM of #YouTube video classification. Founder of telco operator.

[Aurélien Geron @aureliengeron]

My new book just came out! 🎉😊
Kindle & e-book available now, print within 1–2 weeks.

You can get it at: https://t.co/OzaSoITBnE
(you'll also find free online content there)

Play with the notebooks at:
https://t.co/XZHbIc3NH9

Hope you'll find it useful!

[Aurélien Geron @aureliengeron]

@baizhanbaibolat @OReillyMedia Thanks Baizhan!
It's so hard to keep up, but also a lot of fun. 😄
I recently published a new edition using PyTorch:
see https://t.co/OzaSoITBnE

[Aurélien Geron @aureliengeron]

@HanchungLee Ctrl-C = interrupt
Ctrl-D = exit
Comes from Unix/POSIX terminal systems, and works in many shells: Bash, Zsh, Python, Node.js, Ruby, PHP, Go, PostgreSQL, MySQL / MariaDB, SQLite, Docker, ssh, Redis, etc.
I like it: I can interrupt a program without closing the shell by mistake.

[Republicans against Trump @RpsAgainstTrump]

This is INSANE: Trump’s DOJ says the government is now “FOREVER BARRED” from pursuing tax claims against Trump or his family as part of the “settlement.”

The one-page document, signed by Deputy AG Todd Blanche, also blocks claims tied to pending tax returns.

The most corrupt president in American history by far.

[Aurélien Geron @aureliengeron]

@pique90s You can run everything locally using Jupyter if you don't want to use Colab: see the instructions at https://t.co/PyNiqpYaIX
Most notebooks will download datasets, but this only happens once.
Once you have all the code, libraries and datasets, everything should be fully local. 👍

[Aurélien Geron @aureliengeron]

My new book just came out! 🎉😊
Kindle & e-book available now, print within 1–2 weeks.

You can get it at: https://t.co/OzaSoITBnE
(you'll also find free online content there)

Play with the notebooks at:
https://t.co/XZHbIc3NH9

Hope you'll find it useful!

[Aurélien Geron @aureliengeron]

@skenzyme @OutSomebod70162 @nattyover Glad you find me charming. 😅 A 1 followed by 44 zeros is a large number. The question is: is it large enough? Well, all dog breeds evolved from wolves in less than 30,000 years. Evolution is fast, because it is NOT random: mutations are random, but survival is not.

[Aurélien Geron @aureliengeron]

@OutSomebod70162 @skenzyme @nattyover The Earth is over 4 500 000 000 years old. There are about 5 000 000 000 000 000 000 000 000 000 000 bacteria on Earth, and they reproduce every few hours. Each new bacteria is an opportunity for the evolution of the flagellum. Evolution had time. Please watch the videos.

[Aurélien Geron @aureliengeron]

@skenzyme @OutSomebod70162 @nattyover The videos show that there's plenty of evidence for evolution, both from the historical record and from direct observation, in particular in domesticated species and in the lab. It's like a crime scene with the killer's DNA, fingerprints, and surveillance footage.

[Aurélien Geron @aureliengeron]

@OutSomebod70162 @nattyover I know it feels that way, but if you watch the Stated Clearly playlist, you'll see that evolution can be immensely creative and beautiful. It's a real eye opener. ♥️

[Aurélien Geron @aureliengeron]

@prsna01 That's very cool! I'd love to see what the final graph looks like. 👍

[Aurélien Geron @aureliengeron]

@duncanreyburn You might want to read The Righteous Mind by J. Haidt as well. He shows that we tend to decide what to believe based on our education and gut feelings rather than logical arguments. We then cherry pick rational arguments to justify these beliefs. No ideal to get to the truth.

[Aurélien Geron @aureliengeron]

@duncanreyburn So basically any philosopher you disagree with is not a decent philosopher, is that it?
Try reading Daniel C. Dennett: his arguments are really quite convincing. For example in Quining Qualia or Consciousness Explained.

[Aurélien Geron @aureliengeron]

@cartright88 Thanks a lot Adam! 🙏

[Lucas Maes @lucasmaes_]

JEPA are finally easy to train end-to-end without any tricks!

Excited to introduce LeWorldModel: a stable, end-to-end JEPA that learns world models directly from pixels, no heuristics.

15M params, 1 GPU, and full planning <1 second.

📑: https://t.co/cpTzgvbTS0

[alphaXiv @askalphaxiv]

Yann LeCun and his team dropped yet another paper!

"V-JEPA 2.1: Unlocking Dense Features in Video Self-Supervised Learning"

In this V-JEPA upgrade, they showed that if you make a video model predict every patch, not just the masked ones AND at multiple layers, they are able to turn vague scene understanding into dense + temporal stable features that actually understands "what is where".

This key insight drove improvements in segmentation, depth, anticipation, and even robot planning.

[Aurélien Geron @aureliengeron]

@KaiLiuClemson Ah yes, you're right, thanks, I'll drop one of them. 👍

[Aurélien Geron @aureliengeron]

If you're interested in image generation, I just added a simple code example using flow matching at the end of notebook 18:
https://t.co/q5mwTp8EPQ

[Aurélien Geron @aureliengeron]

@KaiLiuClemson Yes, there was a typo, it's fixed now, sorry!

[Aurélien Geron @aureliengeron]

@hakalyanpur Oops, that was a typo. It's fixed now, sorry about that.

[Aurélien Geron @aureliengeron]

@nyang20258 Thanks! Yes, the notebook shows how to build and train these models from scratch. It also includes some examples of using a pretrained model using Hugging Face libraries.