Suhas Pai

@piesauce

Tweets about machine learning, , Toronto, election predictions, public transit. ImageNet Roulette classified me as a psycholinguist.

Vrijeme pridruživanja: veljača 2009.

Tweetovi

Blokirali ste korisnika/cu @piesauce

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

  1. proslijedio/la je Tweet
    1. velj

    99 smartphones are transported in a handcart to generate virtual traffic jam in Google Maps. Through this activity, it is possible to turn a green street red which has an impact in the physical world by navigating cars on another route!

    Prikaži ovu nit
    Poništi
  2. 29. sij
    Poništi
  3. 29. sij
    Poništi
  4. proslijedio/la je Tweet
    22. sij

    TIL how the gradients for the MMI objective function are computed. I have now updated my blog post on LF-MMI and the Kaldi chain model to include this (among other things):

    Poništi
  5. 22. sij

    Is there an NLP paper that has intrigued you lately or that you want to learn more about? Please consider presenting it at the AISC NLP Abstract Night on Feb 3 in Toronto. It will be a welcoming low(no)-pressure environment.

    Poništi
  6. proslijedio/la je Tweet
    18. sij

    Tried mixed precision yet? Took 10 min to set up and my model runs almost 2x faster with same results. Vars and grads are still 32 bits so it usually doesn't affect predictive performance. E.g. in TF2, set option and make all input to your layers float16 (data, RNN states, ..):

    Prikaži ovu nit
    Poništi
  7. 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
  8. 12. sij

    If you are in the Toronto area, come drop by tomorrow for a talk on controlled text generation!

    Poništi
  9. 11. sij

    This is fascinating! I am sure the Canadian version looks somewhat similar.

    Poništi
  10. 10. sij

    Wow! CIFAR is organizing a townhall session on Jan. 15 in Toronto for consultations on their recommendations for the Canadian Federal government's AI for Health strategy.

    Poništi
  11. proslijedio/la je Tweet
    10. sij

    Great talk! Explains how to vectorize slow pandas code. Here: replacing .apply when working w conditional statements. Was guilty of using .apply myself a lot recently because I thought of it as elegant. Turns out my old & actually preferred method, numpy.where, is a lot faster!

    Prikaži ovu nit
    Poništi
  12. 9. sij
    Poništi
  13. 8. sij
    Poništi
  14. 6. sij

    The food item I miss the most from the Netherlands is katjang pedis. I have been trying to find them in Toronto from quite some time now but to no avail :-(

    Poništi
  15. proslijedio/la je Tweet
    1. sij

    We present our new year special: “oLMpics - On what Language Model pre-training captures״, , Exploring what symbolic reasoning skills are learned from an LM objective. We introduce 8 oLMpic games and controls for disentangling pre-training from fine-tuning.

    Prikaži ovu nit
    Poništi
  16. 1. sij

    I set out to write down my personal goals for the new decade. It soon ballooned to become a vision document with 110 action items lol. I do recommend writing out decade level goals, I did it the last time and it served me as a lodestar throughout.

    Poništi
  17. 1. sij

    A happy new year to everyone! I hope you have a decade filled with love, joy, wonder, and adventure.

    Poništi
  18. proslijedio/la je Tweet
    31. pro 2019.

    Worth reading about the peak end rule before reading summaries of the 2010’s.

    Poništi
  19. 27. pro 2019.

    What a decade it has been for me personally! From being a person who had never ventured more than 500 kms away from my birthplace to immigrating to two different continents, each time completely reinventing myself, it has been a wild journey!

    Poništi
  20. proslijedio/la je Tweet
    23. pro 2019.

    This is a delightful information extraction “failure” mode illustrating entity linking gone awry (or missing?) in Google’s backend services. In the fat long tail of such fails will keep us on our toes for quite some time to come, despite impressive feats on benchmarks etc

    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:

    ·