Mehrdad Yazdani

@crude2refined

Data science and machine learning researcher, practitioner, and educator. Previously , , , , . ATM: startup 💻📱

Palo Alto, CA
Vrijeme pridruživanja: prosinac 2013.

Tweetovi

Blokirali ste korisnika/cu @crude2refined

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @crude2refined

  1. Prikvačeni tweet
    10. lip 2018.

    Starting a semi-regular mailing list to share useful resources for machine learning my colleagues and I have found in our projects from the lab to startup land. 🧐🤠 Sign up here and be in the know!

    Poništi
  2. 2. velj

    A single estimate of growth rate gives a single curve, but a single estimate doesn’t capture all the uncertainty. If instead we take uncertainty into account, then we have a COLLECTION of curves...& the avg of the collection is different from the single estimate! Great post!

    Poništi
  3. 29. sij

    Remarkable advance in chatbots. Another takeaway from the paper is using neural architecture search to improve (“evolve”) the Transformer architecture.

    Poništi
  4. proslijedio/la je Tweet
    28. sij
    Odgovor korisniku/ci

    “loss.backward() is all you need” 😄🤔

    Poništi
  5. proslijedio/la je Tweet
    21. sij

    I asked a serial founder what the biggest difference is between his (now) 4th company and his first company - “Even when everyone is panicking, I feel calm. I have seen almost everything that goes right and wrong. No one thing will make or break you” When in doubt, stay calm

    Prikaži ovu nit
    Poništi
  6. proslijedio/la je Tweet
    19. sij

    • Read everyday. • Spend time with nature. • Ask questions. • Never stop learning. • Don't pay attention to what others think of you. • Do what interests you the most. • Study hard. • Teach others what you know. • Make mistakes and learn. • It's Okay to not know things!

    Poništi
  7. proslijedio/la je Tweet

    Productivity tip: At the start of the week, write down 2-3 big decisions you need to make that week. Carve out uninterrupted blocks in the morning to do the work to make those decisions, along with scheduling any necessary meetings. End the week with those decisions made!

    Poništi
  8. 19. sij

    There are good methods for dealing w/ different data sources and representation learning can do away w/ manual feature engineering. I have yet to see representation learning deal at scale w/ such diverse set of heterogeneous data! Images and sound is one thing, but ALL this??!

    Poništi
  9. proslijedio/la je Tweet
    16. sij
    Odgovor korisniku/ci

    You have known this for years so I am very curious about what triggered the tweet! ps - Here is my current version of that thought: The term data science exists because academic statisticians and computer scientists don't care enough about applications.

    Poništi
  10. proslijedio/la je Tweet
    15. sij

    Review deadline for CVPR is approaching. I would suggest reviewers to review a paper like you want others to review your paper. We should try to make review process as a constructive activity wherein researchers can improve upon their work instead of getting discouraged.

    Poništi
  11. 15. sij

    Fantastic study and strong validation sets!

    Poništi
  12. proslijedio/la je Tweet
    13. sij

    Can an deep neural network be trained to diagnose low blood sugar from the ECG signal? I wouldn't have guessed this was possible, but here it is, by and team

    Poništi
  13. proslijedio/la je Tweet
    11. sij
    Prikaži ovu nit
    Poništi
  14. proslijedio/la je Tweet
    10. sij

    Very happy to share our latest work accepted at : we prove that a Self-Attention layer can express any CNN layer. 1/5 📄Paper: 🍿Interactive website : 🖥Code: 📝Blog:

    Prikaži ovu nit
    Poništi
  15. proslijedio/la je Tweet
    9. sij
    Poništi
  16. proslijedio/la je Tweet
    7. sij

    We're in -- again 🥳! With we showed that unlike DTW, soft-DTW captures time-shifts quadratically. With an Optimal transport dissimilarity, we define OT for spatio-temporal data. 🐍Code | Paper |

    Poništi
  17. proslijedio/la je Tweet
    5. sij

    Consider: millions of years ago our antecedents gave a massive sacrifice of their left hemisphere. We lost a tremendous amount of short term memory and replaced it with Broca’s, Wernicke & the phonological loop. But why? So we can—talk. Thus chimpanzees can do this—we can’t:

    Prikaži ovu nit
    Poništi
  18. 4. sij

    Was told management goes 3 ways: upward (ie, managing your own boss), downwards (folks under you), and sideways (your,colleagues). These tips apply in all directions. A manager’s manifesto via

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

    How can we get robots to imitate humans? CycleGAN can turn human videos into robot videos, and model-based RL can execute actions that track these videos. w/ Laura Smith, N. Dhawan, M. Zhang,

    Prikaži ovu nit
    Poništi
  20. proslijedio/la je Tweet
    30. pro 2019.

    I agree with this article that AI Dungeon 2 (the GPT-2 powered text-based game of infinite possibilities) is "one of the coolest video game experiments of 2019". You should definitely check it out: . Great work !

    Poništi
  21. 30. pro 2019.

    Disappointed the often cited “Why We Sleep” book is poorly and deceptive work. Editing data out is never taken well.

    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:

    ·