Matthew Honnibal

@honnibal

Computational linguist from Sydney and Berlin. 💫 Author of the NLP tools. 💥 Founder 👉

Berlin, Germany
spacy.io
Joined May 2008
3 Photos and videos Photos and videos

Tweets

You blocked @honnibal

Are you sure you want to view these Tweets? Viewing Tweets won't unblock @honnibal

  1. Pinned Tweet
    3 Oct 2016

    Say hello to , our new digital studio for custom and machine learning models.

    1 reply 19 retweets 65 likes
  2. Retweeted
    7 hours ago

    code (3.5+), experiments and notebooks:

    Self-Normalizing Neural Networks. Technique to get activations to auto-converge towards zero mean and unit variance.
    1 reply 11 retweets 28 likes
  3. Retweeted
    16 hours ago
    Replying to

    This is the magic SELU activation function.

    7 replies 53 retweets 171 likes
  4. Retweeted
    19 hours ago

    The "Whoa are you serious" award for an Appendix goes to "Self-Normalizing Neural Networks" proposes "selu" nonlin

    16 replies 82 retweets 262 likes
  5. Retweeted
    Jun 8

    17 (!!) Phd and post-doc positions available at newly funded SFB on linguistic variability and stability

    13 retweets 10 likes
  6. Jun 9
    Replying to

    We just need to make it less arcane. Esp. good would be Cython-CUDA. We write CUDA in strings, like beasts. We can do better. 5/ /thread

    1 retweet 4 likes
  7. Jun 9
    Replying to

    JIT compilation is no solution. The programmer *must* make decisions. The C extension story in Python is great. Let's do more of this. 4/

    3 likes
  8. Jun 9
    Replying to

    Programming *entirely* against flat memory buffers is no way to live. The performancish dialect of Python is very hard to read and write 3/

    1 like
  9. Jun 9
    Replying to

    You can't plan out memory in pure Python. How much memory for list of 10**6 ints take? I dunno -- a lot? with numpy, 4mb. Good. But! 3/

    1 reply 4 likes
  10. Jun 9
    Replying to

    Many problems are much easier if you simply sit down and plan out how your data will be laid out in memory. This is not hard! 2/

    3 retweets 5 likes
  11. Jun 9
    Replying to

    So, how can AI in Python be even better? First: accept and embrace native extensions (esp. Cython) Currently seen as last resort. 1/

    2 replies 3 retweets 13 likes
  12. Jun 9

    Excited to be giving a keynote at PyCon Israel next week 🌵 Here's what I'll be talking about:

    7 replies 16 retweets 56 likes
  13. Retweeted
    Jun 8

    So I took some time and wrote this medium post about this language generation paper.

    Replying to @yoavgo
    Maybe it's worth considering writing a blog post with some analysis in case you feel that strongly about it?
    42 replies 272 retweets 531 likes
  14. Retweeted
    Jun 8

    New mini library: termynal.js – lightweight & modern terminal animations with async/await, CSS & plain HTML API. ⬛️

    4 replies 10 retweets 43 likes
  15. Retweeted
    Jun 6

    Gran noticia! , tu librería de , habla ESPAÑOL!! 🗣️🇪🇸🎉 un lujo colaborar con los maestros

    spaCy now speaks Spanish! 🎉 Available for v1.8+, and coming soon for v2.0 alpha. For more info, see the models docs:
    2 replies 36 retweets 30 likes
  16. Retweeted
    Jun 6

    spaCy now speaks Spanish! 🎉 Available for v1.8+, and coming soon for v2.0 alpha. For more info, see the models docs:

    3 replies 22 retweets 46 likes
  17. Retweeted
    Jun 5

    Call for Machine Learning research mentor. July 3-14. Ghana MISE research foundation. For more details please contact me negar@elementai.com

    10 retweets 12 likes
  18. Retweeted
    Jun 5

    Excited to announce spaCy v2.0.0 alpha! New: neural model (15 MB), GPU, Pickle & many training & API improvements.

    3 replies 78 retweets 163 likes
  19. May 31
    Replying to

    1296 combinations! No wonder this is taking so long...

    1 reply
  20. May 31

    Train {parser,tagger,ner} on {py2,py3} {cpu,gpu} {osx,*nix,win} save to {bytes,disk,pickle}. Test on {py2,py3} {cpu,gpu} {osx,*nux,win}.

    1 reply 1 retweet 9 likes
  21. Retweeted
    May 29

    Sleep well tonight.

    0:13
    This is how Tickle Me Elmo looks with his fur removed
    34 replies 195 retweets 595 likes

@honnibal hasn't Tweeted yet.

Loading seems to be taking a while.

Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.

    You may also like

    ·