Hillary Sanders

@hillarymsanders

Senior data scientist at Sophos. Machine-learner, painter, less-wronger. Book: Malware Data Science Tweets reflect my personal views. There it is!

Portland, OR
Vrijeme pridruživanja: siječanj 2015.

Tweetovi

Blokirali ste korisnika/cu @hillarymsanders

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

  1. 4. velj

    Elastic Weight Consolidation does indeed help our PE model avoid catastrophic forgetting! A little bit better than L2 regularization. I'm not convinced it's worth it for standard production releases (keep em' simple!), but it's very cool to see stuff like this work!

    Poništi
  2. 3. velj

    This is a cool plot I made demonstrating mild catastrophic forgetting in our PE model, using ≈toy datasets of 4m/month: Compare the dark and light teal lines to see how fully retraining a model on all past data (9 months) compares to sequential fine-tuning from a base model.

    Poništi
  3. 3. velj

    Informative, noting-when-we-don't-know (👏) update article on the coronavirus that my friend wrote.

    Poništi
  4. 16. sij

    I feel like when I read a whitepaper in blog-format, it's usually quite understandable, but then if I switch to reading the SAME THING in the whitepaper format, my brain gets intimidated and I feel like I'm probably misunderstanding things. 😂😂😂 Does this happen to you, too?

    Poništi
  5. 23. pro 2019.

    Went to a packed high school reunion tonight. Feeling dearly grateful I work from home and mostly just play with silent programs all day for work 😱😍😇.

    Poništi
  6. 17. pro 2019.

    Crushin bugs like a thug who likes to give hugs, and is actually... super non-violent. I mean ok, I was just debugging. Whatever, go away!

    Poništi
  7. 5. pro 2019.

    One way to interpret this plot is that when you're young, stereotypes (read: bayesian priors) affect your perception of your own ability more, because you have less data on your true ability. As women get older, they get more data, weakening those bayesian priors (stereotypes).

    Poništi
  8. 24. stu 2019.

    More reason to be epistemically paranoid! A popular "Why We Sleep" book authored by a professor of neuroscience and psychology at UC Berkeley (yay, Alma Mater...) is *riddled* with factual errors:

    Poništi
  9. 20. stu 2019.

    AI over the last 7 months - TLDR: large costs, small gains, and an unending deluge of PR exaggerations.

    Poništi
  10. 20. stu 2019.

    Being rejected from conferences is like x% sadface, (100-x)% relief you don't have to write up a lil' paper and give a talk. Today was edging around x=55 for my RSA rejection 😂. Next time RSA!

    Poništi
  11. 11. stu 2019.

    Well. It's usually true. As I'm sure it was back then.

    Prikaži ovu nit
    Poništi
  12. 11. stu 2019.

    Remembering this can make it easier to identify bad shit in our current society, and can possibly (?) make it easier to turn down the hate on people who we think are wrong.

    Prikaži ovu nit
    Poništi
  13. 11. stu 2019.

    Good to remember: positions that today we consider undisputedly bad, were made with (strong) steel-man arguments, not (weak) straw-man arguments back in the day. Also, the antiheros came - at least in part - from a place of warmth. The same is true today.

    Prikaži ovu nit
    Poništi
  14. 28. lis 2019.

    , the place I go to read about [tech] news and community responses, with *just* a splash of the Red Pill.

    Poništi
  15. proslijedio/la je Tweet
    23. ruj 2019.

    Reminder that all of the royalties make from Malware Data Science go to the in support of combating climate change and promoting science driven environmental policy. This will total $20k after our next royalty payment. Thank you to our readers.

    Poništi
  16. 4. ruj 2019.

    Often, Bob states unfair sweeping generalizations about people to signal to Bob's in-group that Bob is a Member 😎. Also, so cathartic! Taking this literally, people not in Bob's in-group feel defensive and move further away on the political spectrum from Bob. Don't be like Bob

    Poništi
  17. proslijedio/la je Tweet
    28. kol 2019.

    Are journalists who cover terrorist attacks doing something morally impermissible?

    Poništi
  18. 29. kol 2019.

    If you feel upset / betrayed because you think someone is signaling that they're not a member of your belief group (e.g. states opinion x, which is meh, but you think x is associated with Evil-Group X) it's almost always a good idea to let those feelings go. Just focus on x.

    Poništi
  19. 30. srp 2019.

    One can criticize our current tax system, and that's great. We should tax the super-rich more! But criticizing billionaires for donating their own money? We should be applauding and encouraging those actions, not disincentivizing them with a barrage of negative press.

    Prikaži ovu nit
    Poništi
  20. 30. srp 2019.

    The system that creates billionaires is bad. But the trend of criticizing billionaires for donating their money to charity instead of blowing it on yachts, mansions and their dynasty is insane and harmful. Billionaire philanthropy is great.

    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:

    ·