Jeremy Howard

@jeremyphoward

Deep learning R&D and education. Founder: ; Faculty: ; Chair ; // Previously - CEO: Enlitic; President: Kaggle; CEO Fastmail

San Francisco
Vrijeme pridruživanja: kolovoz 2010.

Tweetovi

Blokirali ste korisnika/cu @jeremyphoward

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @jeremyphoward

  1. Prikvačeni tweet
    28. lip 2019.

    Learn: - Practical Deep Learning - Deep Learning Foundations - Computational Linear Algebra - Intro Machine Learning

    Poništi
  2. proslijedio/la je Tweet
    prije 2 sata

    Tesla researchers authors NNs in python land and rewrite with a bare C++ implementation when deploying. This feels like a failure of our tooling / infrastructure.

    Poništi
  3. proslijedio/la je Tweet
    prije 7 sati

    In January, , , and I ran a short class at on topics we think are missing in most CS programs — tools we use every day that everyone should know, like bash, git, vim, and tmux. And now the lecture notes and videos are online!

    Poništi
  4. prije 2 sata

    Seems like an odd choice. Over $1m of compute costs to train, but Google didn't use the best available version of the tool they compared to.

    Poništi
  5. proslijedio/la je Tweet
    prije 2 sata

    I borrowed 's idea and am having my data ethics students keep a group spreadsheet of current data ethics news articles. Some articles from the past week:

    Article Title
Have a Search Warrant for Data? Google Wants You to Pay
The 'Illinois Artificial Intelligence Video Interview Act' is a real law. Here's why it may be coming to a job application near year.
Leaked Documents Expose the Secretive Market for Your Web Browsing Data
Data privacy must be seen as human right: Satya Nadella
India likely to force Facebook, WhatsApp to identify the originator of messages
Singapor
    The battle for ethical AI at the world’s biggest machine-learning conference
"One year inside Trump's monumental Facebook campaign
"
Larry Magid: Unintended consequences of well-meaning innovations and laws
Avast shuts down marketing analytics subsidiary Jumpshot amid controversy over selling user data
40 groups have called for a US moratorium on facial recognition technology
"The big data of the goverment: The detai
    Poništi
  6. prije 4 sata

    "We achieved an overall accuracy of 94.5%, more than 4.5% of an increase on the previous state-of-the-art"; classifying the patterns in 18,577 Scanning Electron Microscope images

    Poništi
  7. proslijedio/la je Tweet
    2. velj
    Odgovor korisniku/ci

    A PhD is definitely not required. All that matters is a deep understanding of AI & ability to implement NNs in a way that is actually useful (latter point is what’s truly hard). Don’t care if you even graduated high school.

    Poništi
  8. proslijedio/la je Tweet
    prije 7 sati

    Introducing Swift Crypto, a cross-platform package that provides an open source implementation of the platform independent crypto APIs from Apple’s CryptoKit library:

    Poništi
  9. proslijedio/la je Tweet
    prije 6 sati
    Odgovor korisnicima

    If using EfficientNets in , definitely make sure you're using a jit'd Swish function with custom 'memory efficient' autograd. Will help bump up the batch sizes a little and squeeze out a few more img/sec.

    Poništi
  10. proslijedio/la je Tweet
    prije 7 sati

    Check out our new tutorial about how to perform efficient data augmentation in videos with for the Deepfake Detection Challenge

    Poništi
  11. proslijedio/la je Tweet
    3. velj

    Session 02 is Tuesday! Join for cancer genomics live-coding as Dr. (Stanford Medicine) and ( ) experiment with visual GPU UMAP for longitudinal analysis over thousands of cancer patients in GENIE / !

    Poništi
  12. prije 18 sati

    I'll be sure to include %debug, since I know that's a favorite! And I'll go thru this list, so no need to mention stuff from here: )

    Prikaži ovu nit
    Poništi
  13. prije 18 sati

    During our video lessons I do a lot of little asides about debugging, s/w eng, ethics, etc. I want our upcoming book to include these too! What tidbits from the courses should we be sure to include in the book?

    Prikaži ovu nit
    Poništi
  14. proslijedio/la je Tweet
    prije 22 sata

    in search of personal analog style in the age of styleGAN paintings painfully aware of how i became gan-conscious while casually sketching, every drawing is a training material, every line - what will the model make out out of it?..

    Prikaži ovu nit
    Poništi
  15. prije 21 sat

    This repo is full of amazing awesomeness. I don't know of anything else like it. Independent refactored carefully tested implementations of modern CNNs

    Poništi
  16. proslijedio/la je Tweet
    prije 21 sat
    Odgovor korisnicima

    I recommend reading the forward I wrote for - Rust and Swift have very very very different goals, it is not just about target market, type system or syntax. In any case, I am a huge fan of Rust, its community, and many aspects of its design!

    Poništi
  17. proslijedio/la je Tweet
    Odgovor korisnicima

    A relu is a diode and was invented in 1874. They had valve tube diodes for experiments when I was studying! The transfer function suspiciously looks like relu. :-)

    Poništi
  18. prije 24 sata
    Poništi
  19. proslijedio/la je Tweet
    3. velj
    Odgovor korisnicima

    IIRC Fukushima's Neocognitron also used the max(0, x) function in its design (to mimic firing frequencies in biological neurons, which must be positive).

    Poništi
  20. proslijedio/la je Tweet
    2. velj

    I’m confused. How did we conclude that the best way to prepare kids for the future is to cluster them into a setting where they are organized by age, into grades, and forced to learn the same things, at the same time and pace, 7 hours a day, 5 days a week, for 12 + years? Huh?

    Poništi
  21. proslijedio/la je Tweet

    Interesting costs/spec for Google's Meena chatbot. By comparison, Mitsuku takes 40Mb of space, needs 4Mb of RAM and runs on a single CPU. As a bonus, I even let people talk to it! 😉

    Prikaži ovu nit
    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·