Michael Carbin

@mcarbin

Assistant Professor in EECS at | Programming Systems | Approximate Computing | Tiny Neural Networks

Cambridge, MA
Vrijeme pridruživanja: rujan 2007.

Tweetovi

Blokirali ste korisnika/cu @mcarbin

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

  1. Prikvačeni tweet
    6. svi 2019.

    Congrats ! I'm very privileged to work with fantastic students like Jonathan.

    Poništi
  2. proslijedio/la je Tweet
    23. sij

    WACI is my favorite part of my favorite conference, . We want your sprawling plans for world domination! Write up a one-paragraph abstract now and finish your paper & talk later. It’s low risk, sky-high reward. You can’t afford not to WACI.

    Prikaži ovu nit
    Poništi
  3. proslijedio/la je Tweet
    21. sij

    We just sent out an RFC for implementing our timing framework BHive () inside LLVM-exegesis. Please provide feedback to improve our design. ()

    Poništi
  4. proslijedio/la je Tweet
    14. sij

    How Fast Does This Code Run? MIT researchers have trained a neural network to tell you how fast any code that you present it with will run. Sounds fun but why do we need it?

    Poništi
  5. proslijedio/la je Tweet
    9. sij

    Are you an early-stage PhD student in computer architecture? If so, consider submitting your ongoing work to the YArch mentoring workshop co-located with 2020 (Lausanne, Switzerland). Here is more:

    Prikaži ovu nit
    Poništi
  6. 10. sij
    Poništi
  7. proslijedio/la je Tweet
    2. sij

    People tweet all sorts of things about real numbers, especially when it comes to computation. I've studied the topic for years, and I often disagree with such statements. Let me address just one such statement: "A computer cannot represent all reals, only the computable ones."

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

    Congrats to and on an oral for "Comparing Fine-tuning and Rewinding in Neural Network Pruning." You get better accuracy if you replace fine-tuning with lottery ticket rewinding in the pruning workflow.

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

    If medicine were like computer security, researchers would be celebrated for creating superbugs that defeat existing antibiotics.

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

    The next deadline falls on a Saturday evening. What a terrible message this sends to our community about inclusion and work/life balance. I wrote to about it, and I hope others do, as well.

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

    When I first became obsessed with program understanding over a decade ago, I used to tell people that the collective "we" need to understand software because it's going to run our cars and lives. I LOVE IT WHEN MY HYPOTHETICAL MOTIVATIONS BECOME REAL!! 🏆

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

    Does anyone know of any physical CPUs that adhere to some relatively simple model? E.g., a simple in-order processor, with some ability to profile execution times?

    Prikaži ovu nit
    Poništi
  13. proslijedio/la je Tweet

    also this is a good time to say a huge THANK YOU to all the people who wrote letter for me throughout the past 20+ years ... for college admissions, for Ph.D. admissions, for the 2930434 jobs i had post-Ph.D., for my academic promotion files, etc. THANK YOU!

    Prikaži ovu nit
    Poništi
  14. 15. pro 2019.

    Lots of pain to systematically profile basic blocks, now hopefully you don’t have to. COMMIT () and PSG collaboration, led by and with many more great students (e.g., , ). Spanner and Dremel analysis courtesy of .

    Poništi
  15. 14. pro 2019.
    Poništi
  16. 14. pro 2019.

    Another step forward in our understanding of the mechanisms behind lottery tickets.

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

    Implementing probabilistic systems means there will be approximations. Can we control the error when we compose multiple approximate programs that are potentially nested? We give some guarantees for the error induced by MCMC based approximations. Work with .

    Prikaži ovu nit
    Poništi
  18. proslijedio/la je Tweet
    14. pro 2019.

    Visit our poster from 12:00-2pm at the Program Transformations for ML workshop, Room West 114.

    Poništi
  19. proslijedio/la je Tweet
    13. pro 2019.

    Come chat about using Tiramisu for sparse neural networks and LSTMs! The afternoon poster session is 3:30-4:30, in rooms 11 and 12 at the MLSys Workshop.

    Poništi
  20. proslijedio/la je Tweet

    🎉🎉Excited to announce our panel 🎉🎉 Come see Yoshua Bengio, Joelle Pineau, , , and moderated by discuss retrospectives today at 11:15 , West 114+115

    Prikaži ovu nit
    Poništi
  21. 13. pro 2019.

    Learn about Tiramisu and how we’ve been compiling Lottery Tickets for better performance. Big team at with Riyadh in the lead and including my group’s students, and

    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:

    ·