Edward Dixon

@EdwardDixon3

Intel Data Scientist. Interest in stems from hope that someday a robot will iron my shirts, co-author of 'Demystifying AI for the Enterprise'.

Vrijeme pridruživanja: siječanj 2012.

Tweetovi

Blokirali ste korisnika/cu @EdwardDixon3

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @EdwardDixon3

  1. Prikvačeni tweet
    27. stu 2018.

    Just how easily can your model-as-a-service get pirated? And, just how fat are those AWS Rekognition profit margins anyway? Answers to these and more in our new article (with and )

    Poništi
  2. proslijedio/la je Tweet

    3. The number of suspect cases lists here kind of grabs me by the throat. Hard to see that this horse could be coaxed back into the barn.

    Prikaži ovu nit
    Poništi
  3. 31. sij

    Really stunning data efficiency, does very well with small training sets.

    Prikaži ovu nit
    Poništi
  4. 31. sij

    At this point 's newsletter might as well go to daily publication, and with work like this, maybe (semi-)automatically generated newsletters are not complete fantasy?

    Prikaži ovu nit
    Poništi
  5. 31. sij

    Great interview, fascinating stuff on the power of and limitations to and . Your sort of thing I believe.

    Poništi
  6. 29. sij

    A great paper deserves a great review video, and is in fine form here. Lots to interest folk, regardless of interest in . Come for the "Hayvard" joke, stay for the awesome power of BPE and the evolved transformer architecture.

    Poništi
  7. 29. sij

    Just for context, the amount of anti-matter in one solar flare would yield about 2X global annual demand for electricity (!). Scoop it up from 3 flares & you've got 1.5kg, enough to exceed total global energy demand. Now we just need to figure out storage & distribution...

    Poništi
  8. 28. sij

    A stunning paper from on a chatbot that manages a pretty neat joke that really does seem to be original (see below, "Horses go to Hayvard").

    Poništi
  9. 28. sij

    I had always wondered in a vague way about the mathematical underpinnings of maze generation. has a great post about the maths of generating them, and of solving them too!

    Poništi
  10. 28. sij

    It has really helped me, and friends I referred to it also found it increased their productivity. Great tool for folk, delighted to see this news.

    Poništi
  11. 27. sij

    Playing with and 's collected works, I used a powered library from to make his sentences searchable by meaning (not spelling). Looking for "Where is the CEO?", Sentence-BERT found these...

    Poništi
  12. 26. sij

    An exceptional interview, both inspiring and highly informative.

    Poništi
  13. 24. sij

    Sentence-BERT paper is on Arxiv , and the sentence-transformers package repo is on . Really nice work!

    Prikaži ovu nit
    Poništi
  14. 24. sij

    Experimenting with Sentence-BERT from , powered by . Queried customer service ticket dataset w/ "computer unresponsive", great results! I made a notebook so you can try it too recommended.

    Prikaži ovu nit
    Poništi
  15. 18. sij

    Brain science, mice and all wrapped up in one little thread... nice!

    Poništi
  16. proslijedio/la je Tweet
    18. sij

    Still programming tonight. Am in a gnarly bit of code...I have John Cage's 4'33" on a loop so that I can concentrate.

    Poništi
  17. proslijedio/la je Tweet
    17. sij

    If you want to know who I am and how my voice was crafted - pease read this:

    Poništi
  18. proslijedio/la je Tweet
    16. sij

    A belated blog post for our BERTology EMNLP paper (by Olga Kovaleva, Alexey Romanov, yours truly and ). My favorite experiment in this work is showing that for most GLUE tasks BERT works pretty well even *without pre-training*!

    Prikaži ovu nit
    Poništi
  19. 16. sij

    Twitter is the best Twitter! A tweet surfaces a paper about helping models cope with rotation and a 3-timezone thread surfaces a connection between 2 mathematical fields and really important physics. Wow!

    Poništi
  20. 15. sij

    Last night my wife & I were discussing whether Eldest Daughter should be encouraged to take up a musical instrument, and this morning helpfully posts this! The paper is worth reading just for the methodology section alone, really excellent.

    Poništi
  21. 14. sij

    is getting ridiculously powerful. Note the results versus Mathematica! Also the huge and ingeniously generated dataset. your kind of thing I believe.

    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:

    ·