Been Kim

@_beenkim

Research Scientist at Google Brain, PhD from MIT on Interpretable Machine Learning

Mountain view, CA
Vrijeme pridruživanja: kolovoz 2011.

Tweetovi

Blokirali ste korisnika/cu @_beenkim

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

  1. Prikvačeni tweet
    7. svi 2019.

    I'm incredibly excited and honored that talked about our work TCAV () today at his keynote! 😀

    Poništi
  2. proslijedio/la je Tweet
    3. sij

    . of Brain strives to make humans empowered by . She gave ICML tutorial in 2017, CVPR and MLSS at . Her work was part of CEO's keynote 2019.

    Poništi
  3. proslijedio/la je Tweet
    14. pro 2019.

    You're at NeurIPS and you want to chat about non-work related stuff? Send me a message and we'll try to meet at the banquet.

    Poništi
  4. proslijedio/la je Tweet
    14. pro 2019.

    Excited for ’s talk on our ML for chip placement work that , Jeff, and I started in 2018. This is one of the first steps in learned ASIC Design. Look forward to sharing more soon. So grateful to all our amazing collaborators in Google Brain and Platforms!

    Poništi
  5. proslijedio/la je Tweet
    13. pro 2019.
    Prikaži ovu nit
    Poništi
  6. proslijedio/la je Tweet

    . of evaluates attribution methods by placing dog in various landscapes, which shouldn’t affect scene label. On same dataset, can train different models w/ object vs scene labels & measure model contrast as proxy for relative influence

    Poništi
  7. 12. pro 2019.

    Excited for the best part of party- workshops! Tomorrow, I'll be giving a keynote and part of panels at human centric ml (west 223) at 4pm, a panel at minding the gap workshop (east 8+15 (?)) at 11:30am! Sat: another fun panel at SEDL workshop 4:15

    Poništi
  8. 12. pro 2019.

    I meant poster #86!!:)

    Prikaži ovu nit
    Poništi
  9. 12. pro 2019.

    Poster today at 10:45am! #83 "Toward automatic conference base explanations"- work with my awesome intern , and two awesome jamess - James Zou. And James Wexler

    Prikaži ovu nit
    Poništi
  10. proslijedio/la je Tweet
    9. pro 2019.
    Poništi
  11. 10. pro 2019.

    Today at 10:45am Poster #172 "Visualizing and Measuring the Geometry of BERT" and #159 "A Benchmark for Interpretability Methods in Deep Neural Networks" - see you there!

    Poništi
  12. proslijedio/la je Tweet
    9. pro 2019.

    I'm happy to share my comments on the climate for men from my talk:

    Poništi
  13. 10. pro 2019.
    Poništi
  14. proslijedio/la je Tweet
    8. pro 2019.

    Congrats to for winning the Outstanding New Directions Paper Award at ! Come see his talk this Tuesday, 9/10, at 10:05am in West Exhibition Hall C + B3. Paper available here:

    Prikaži ovu nit
    Poništi
  15. 9. pro 2019.

    Devavrat Shah tutorial on Synthetic control - bringing it down to ML language - matrices and tensors!

    Poništi
  16. proslijedio/la je Tweet
    8. pro 2019.

    Then, she reminded that commonly used interpretation methods such as saliency maps (Integrated Gradients, Guided Backprop, etc.) are not reliable (really cool 2018paper by et al. )

    Prikaži ovu nit
    Poništi
  17. proslijedio/la je Tweet
    8. pro 2019.

    A very interesting talk by at where she talked about Testing with Concept Activation Vectors (TCAV). For diabetic retinopathy, when a model is less accurate for certain classes, TCAV shows it is also inconsistent with the doctor's knowledge.

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

    Walk into and am figuring where I want to go first and hear the magnificent talking about her work on interpretable . 😀

    Been Kim speaking on stage
    Poništi
  19. proslijedio/la je Tweet
    7. pro 2019.

    This year, as part of the meetup experiment, we have a Discord server with text and audio channels. Feel free to join to discuss NeurIPS papers even if you're not in Vancouver:

    Prikaži ovu nit
    Poništi
  20. 7. pro 2019.

    I'll be also giving a talk at Neurips Expo as a Google speaker, and a keynote at workshop. I'll also be at fun panels at "Mind the gap" workshop (what a great name btw) (Friday) and "Science meets Engineering of Deep Learning" workshop (Saturday)! Get the party STARTED!

    Prikaži ovu nit
    Poništi
  21. 7. pro 2019.

    Hello (aka yearly reunion w/ friends), three accepted papers this year - "Visualizing and Measuring the Geometry of BERT" "A Benchmark for Interpretability Methods in Deep Neural Networks" Tuesday "Visualizing and Measuring the Geometry of BERT" Thursday!

    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:

    ·