Max Christ

[Max Christ @MaxBenChrist]

It's been quite some time since the last tsfresh release. So, we released tsfresh 0.12.0 yesterday. It contains a group of bug fixes and added support for index based features. More details here
https://t.co/PPlW7O6mjN

[Giannis Daras @giannis_daras]

DALLE-2 has a secret language.
"Apoploe vesrreaitais" means birds.
"Contarra ccetnxniams luryca tanniounons" means bugs or pests.

The prompt: "Apoploe vesrreaitais eating Contarra ccetnxniams luryca tanniounons" gives images of birds eating bugs.

A thread (1/n)🧵

[Max Christ @MaxBenChrist]

For juniors, deep learning on time series with keras consists of like 90% being confused about the tensor dimensions

[Max Christ @MaxBenChrist]

Der @airbnb_de twitter Bot konnte auch nicht weiterhelfen. Ich bin unzufrieden über diesen UX Fauxpas, wenn die Unterkunft kein Internet hat, dann sollte das AirBnB Frontend nicht das Wlan Symbol mit der Formulierung "Ausstattung für den Alltag" verwenden.

[Max Christ @MaxBenChrist]

Hey @airbnb_de, warum nutzt Ihr denn das WLAN Symbol für Unterkünfte, die gar kein Internet haben? Bin in die Falle getappt, hat mich eine schöne Storno Gebühr gekostet. Eure Service Hotline Mitarbeiter waren nett, konnten aber nicht helfen. Ärgerlich

[Max Christ @MaxBenChrist]

Thanks

[Nat Friedman @natfriedman]

We spent the last year working closely with OpenAI to build GitHub Copilot. We've been using it internally for months, and can't wait for you to try it out; it's like a piece of the future teleported back to 2021.

[GitHub @github]

Meet GitHub Copilot - your AI pair programmer. https://t.co/eWPueAXTFt

[Peyman Milanfar @docmilanfar]

Don't let low-order statistics fool you

[Max Christ @MaxBenChrist]

Interested in presenting novel techniques for time series feature extraction?
Then consider handing in a paper for the workshop on Systematic Feature Engineering for Time-Series Data Mining during the IEEE ICDM 2021 in Auckland on December 7, 2021. See https://t.co/C92WwH8onp

[Max Christ @MaxBenChrist]

@ErbStephan It took me some time but I finally managed to read the paper. I enjoyed it, thank you for the recommendation

[Max Christ @MaxBenChrist]

best binary format to save univariate time series? (best in terms of compression and reading speed)
best binary format to save multivariate time series? (not necessarily univariately sampled)

[Max Christ @MaxBenChrist]

@xhochy again, a blog post of you that I thoroughly enjoyed reading. Well done Uwe!

[Uwe L. Korn @xhochy]

Thinking about how to store your trained ML models in Python? I did some exploration of techniques beyond pickle in https://t.co/12aUyHLWTP

[Rich Harris @Rich_Harris]

"If not for cryptocurrency, these services would still be available."

https://t.co/J1NnbSsrH7

[Max Christ @MaxBenChrist]

discussion on why PEP 563, PEP 649 might break pydantic https://t.co/sO2LLh4BF3

[Programming Wisdom @CodeWisdom]

"Sometimes the elegant implementation is a function. Not a method. Not a class. Not a framework. Just a function." – John Carmack

[Max Christ @MaxBenChrist]

TIL about Apache TVM which is a compiler stack for deep learning systems. It is designed to close the gap between the productivity-focused deep learning frameworks, and the performance- and efficiency-focused hardware backends. https://t.co/775OjCjZ3d

[Dave Kennedy @HackingDave]

Wow, I am completely speechless here.

Microsoft really did remove the PoC code from Github.

This is huge, removing a security researchers code from GitHub against their own product and which has already been patched.

This is not good.

https://t.co/yqO7sebCSU

[Max Christ @MaxBenChrist]

@bu2adaptive @xhochy @_NicT_ @onnxai tree-based models will not be able to detect weekday versus weekend patterns on seconds, they would need unlimited cutting points to do that. I am not sure if e.g. deep learning models can learn modulo operations. anyway, it makes sense to move the calculation into preprocessing

[Max Christ @MaxBenChrist]

What about using video game engines to create images and videos to train machine learning models? This should generate unlimited, labeled data at low costs? Problem: classifier will probably detect cab drivers, merchants and construction workers as NPCs