Benjamin Rubinstein

@bipr

Associate Prof into Machine Learning, Privacy & Security. PhD. Formerly Silicon Valley.

Melbourne, Australia
Дата приєднання: лютий 2009

Твіти

Ви заблокували @bipr

Ви впевнені, що хочете переглянути ці твіти? Перегляд твітів не розблокує профіль @bipr

  1. Закріплений твіт
    18 лип. 2017

    Introducing diffpriv: a package for easy differential privacy in R.

    Скасувати
  2. ретвітнув(ла)
    11 годин тому

    Selection effects are interesting. Universities like to brag about how small their classes are. Current students often find it hard to believe these stats. "The average class size is 20? No way! In three years, all but 2 of my classes have had at least 100 students." 1/n

    Показати цей потік
    Скасувати
  3. ретвітнув(ла)
    23 груд.
    Скасувати
  4. ретвітнув(ла)
    22 груд.

    NeurIPS: "Not a single motivated PhD student should be denied entry while spectators attend". Excellent piece by Zachary Lipton

    Скасувати
  5. ретвітнув(ла)
    23 груд.

    Happy to report that our paper was accepted at AISTATS'19: RAPTOR goes to Japan! (cc: )

    Показати цей потік
    Скасувати
  6. 22 груд.
    Скасувати
  7. ретвітнув(ла)
    19 груд.

    Australia cuts research funding for universities. Funding for research and development is at its lowest point in 40 years—even as Australia forecasts a budget surplus of AU$4.1 billion in 2019-2020

    Скасувати
  8. 21 груд.

    And some Adversarial ML for 'writing' papers by gradient ascent ;)

    Скасувати
  9. ретвітнув(ла)
    21 груд.

    I am a researcher and I know that university research changes lives. Go to or post on to stand against funding cuts

    Скасувати
  10. ретвітнув(ла)
    20 груд.

    (1/6) I just read in a Nature paper that the authors used "101 CPU-core years of computation (..) within the Google data centres" What could be the carbon footprint of such a study?

    Показати цей потік
    Скасувати
  11. ретвітнув(ла)
    20 груд.

    Leaking information through silence is sometimes a feature rather than a bug. That's the idea behind a warrant canary. An entity that expects to be served a secret subpoena posts a regularly updated notice stating that it has NOT been served such an order.

    Показати цей потік
    Скасувати
  12. ретвітнув(ла)
    18 груд.

    "By complying with requests to selectively obscure military facilities, the mapping service has actually revealed their precise locations, perimeters, and potential function to anyone curious enough to find them all."

    Скасувати
  13. ретвітнув(ла)
    18 груд.

    Despite promising users control over how their data informs ad experience, exercising Facebook’s location controls doesn’t stop location-based ad targeting

    Скасувати
  14. ретвітнув(ла)
    17 груд.

    Crisp piece from explaining why data ownership sounds good, but is really a red herring. Rights is a better framework, especially since many of the protections that matter are *after* we've traded or given away what we "own."

    Скасувати
  15. ретвітнув(ла)
    18 груд.

    Politics is about choices. Cutting $130m a year from research funding is a poor choice: MYEFO rips A$130 million per year from research funding despite budget surplus

    Скасувати
  16. ретвітнув(ла)
    17 груд.

    “Even before these cuts, Australia’s Government spending on R&D was forecast to plunge to a four-decade low of half a per cent of GDP this year — setting loud alarm bells ringing.”

    Скасувати
  17. ретвітнув(ла)
    17 груд.

    Liberal cuts to research harm all Australians, will cost jobs and sacrifice our future. With Australian R&D at the lowest level recording since 2004-5 this cut is shortsighted and shows this is a government which has its priorities all wrong

    Скасувати
  18. ретвітнув(ла)
    17 груд.

    Not for our scientists. Funding cut by $328.5 million over the next four years.

    Скасувати
  19. 16 груд.
    Скасувати
  20. ретвітнув(ла)
    14 груд.

    A couple days ago another team asked me to speak about Bayesian data analysis. I decided that instead of doing a nuts/bolts of how to fit/use Bayesian models, I would describe "Bayesian analysis: The Good Parts". <potentially controversial thread>

    Показати цей потік
    Скасувати
  21. ретвітнув(ла)
    14 груд.
    Скасувати

Схоже, завантаження займе трохи часу.

Можливо, Твіттер перенавантажено або виникли тимчасові труднощі. Спробуйте ще раз або дізнайтеся більше про стан Твіттера.

    Вам також може сподобатись

    ·