Gaurav

@singh_gaurav

The Netherlands
Vrijeme pridruživanja: srpanj 2009.
4 fotografije ili videozapisa Fotografije i videozapisi

Tweetovi

Blokirali ste korisnika/cu @singh_gaurav

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

  1. Prikvačeni tweet
    3. ruj 2019.

    It appears to me that much of human learning at a high level is overfitting to a hypothesis followed by bias correction based on negative examples. Results obtained from such learning are not without flaws and depend greatly on how well the negative examples are used

    1 reply
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  2. proslijedio/la je Tweet
    28. sij

    New paper: Towards a Human-like Open-Domain Chatbot. Key takeaways: 1. "Perplexity is all a chatbot needs" ;) 2. We're getting closer to a high-quality chatbot that can chat about anything Paper: Blog:

    i
    28 replies 629 proslijeđenih tweetova 1.787 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  3. proslijedio/la je Tweet
    25. pro 2019.

    Comparing this to the human brain might be a bit of a stretch but nevertheless fascinating development!

    4 proslijeđena tweeta 8 korisnika označava da im se sviđa
    Poništi
  4. 17. pro 2019.

    Open minds

    1 korisnik označava da mu se sviđa
    Poništi
  5. proslijedio/la je Tweet
    15. stu 2019.

    Australia vs Pluto

    316 replies 3.228 proslijeđenih tweetova 25.997 korisnika označava da im se sviđa
    Poništi
  6. proslijedio/la je Tweet
    14. stu 2019.

    In early days of my career I made a postdoc's code go ~50 times faster. He promised co-authorship on the paper since his method was now viable. Months later, paper came out without my name. Why? Supervisor saw I only changed ~20 lines out of ~1000 and said was not significant.

    Number of lines of code written is not a measure of your value to a project.
    55 replies 568 proslijeđenih tweetova 2.616 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  7. proslijedio/la je Tweet
    24. lis 2019.

    Our Team Scatter fought hard, but it wasn't meant to be :( Nevertheless, we're very proud of what they achieved during Team Scatter, thank you for 3 years of hard work, focus and dedication!

    The elimination match in Round 5 ended in a tie, but after the tiebreaker, claimed victory over Team SCATTER!
    2 proslijeđena tweeta 16 korisnika označava da im se sviđa
    Poništi
  8. proslijedio/la je Tweet
    20. lis 2019.

    If you have never seen a bombardier beetle in action here is a great video. Those ‘pops’ are literally explosions that generate the hot liquid/gas being sprayed in defence.

    0:04
    ミイデラゴミムシの噴射！ 腹端をあらゆる方向に向けて自由に噴射できます。高温なので、これに当たると大変です！ ぜひ音を出してみてください！プシュプシュという音から凄さがわかります！
    Prikaži ovu nit
    145 proslijeđenih tweetova 454 korisnika označavaju da im se sviđa
    Poništi
  9. proslijedio/la je Tweet
    17. lis 2019.

    Our very own explaining that understanding recovery in sports will be the key to improving performance. Will closely monitor the progress of this work ...

    1:51
    Today's film release at about rowing, explains how we monitor the recovery of a whole team and develop algorithms that can be attuned individually.
    4 proslijeđena tweeta 8 korisnika označava da im se sviđa
    Poništi
  10. 16. ruj 2019.

    Would like to hear thoughts on this:

    1 reply
    Poništi
  11. proslijedio/la je Tweet
    9. ruj 2019.

    Need help . Parents are stuck at NRT and even though weather has cleared, ANA has cancelled to DEL and staff has simply left! about 100 passengers are stranded with no Visa, no outbound flight; many without much cash and visa.

    3 proslijeđena tweeta
    Poništi
  12. proslijedio/la je Tweet
    9. ruj 2019.

    Hey , terrible management by your ground staff at NRT. My parents are stuck at the airport with no information and no service. If the flight is cancelled, at least provide accom and rebook. What do you expect international passengers with no JP visa to do ?

    2 proslijeđena tweeta
    Poništi
  13. 3. ruj 2019.

    Very nice review of pruning in neural networks!

    Our Researcher shares his thoughts on sparsity in neural networks by examining the topic of pruning in several different papers.
    3 proslijeđena tweeta 6 korisnika označava da im se sviđa
    Poništi
  14. 3. ruj 2019.

    From this perspective, focus of "AGI" should be on more efficient implementation of architectures that can generalize in presence of large training data. Striving for human-level intelligence is a disservice to progress, we should rather focus on building a better version of us

    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  15. 3. ruj 2019.

    This maybe where AI could (in principle) be better version of us as it can be designed for high bandwidth input and architected generalize sufficiently in presence of large amounts of training.

    1 reply
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  16. 3. ruj 2019.

    Perhaps if this input bandwidth can be expanded with ideas of , we could generalize sufficiently by learning from more training data. But, my internal biases make me believe that there may be fundamental limits as a product of evolution

    1 reply
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  17. 3. ruj 2019.

    But, at macro-level and as a belief-system our learning just cannot generalize in lack of sufficient training examples. Even in presence of a library full of diverse training data, our input bandwidth is far too low to process it all

    1 reply
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  18. 3. ruj 2019.

    The deeper rooted beliefs in the network need serious jolts for us to get rid of. I agree with that humans are not really that smart. We happened to be endowed with a very efficient learning machine that generalizes at sensory-level thanks to million years of evolution

    1 reply
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  19. 3. ruj 2019.

    This is also why optical illusions work, our learning bas overfitted to deal with what is relevant in real world and optical illusions can be viewed as adversarial examples where our learning fails. But, it is sometimes understated how difficult un-learning something is.

    1 reply
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  20. 3. ruj 2019.

    This seems to be in agreement with Max Planck: "A new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with it."

    1 reply
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  21. 3. ruj 2019.

    This can be observed in everyday life where we are constantly learning from limited set of training data with further limitation of processing these training examples with our pre-biased neural networks. This appears in varying degrees in presence/absence of scientific training

    1 reply
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi

@singh_gaurav još nije objavio nijedan Tweet.

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:

    ·