Christopher #AlwaysBeLearning Poptic

@ChrisPoptic

Data scientist, machine learning eng, bike racer. "Here I stand, atoms with consciousness, matter with curiosity. A universe of atoms, an atom in the universe."

ENTP / ENTJ
Vrijeme pridruživanja: svibanj 2009.
30 fotografija ili videozapisa Fotografije i videozapisi

Tweetovi

Blokirali ste korisnika/cu @ChrisPoptic

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

  1. Prikvačeni tweet
    30. ožu 2019.

    "...bear in mind that every man lives only this present time, which is an indivisible point, and that all the rest of his life is either past or it is uncertain. Short then is the time which every man lives, and small the nook of the earth where he lives; and short too the

    1 reply
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  2. proslijedio/la je Tweet
    4. velj

    "How to do machine learning efficiently". There's so much to love about this wonderful article.

    251 proslijeđeni tweet 1.160 korisnika označava da im se sviđa
    Poništi
  3. proslijedio/la je Tweet
    2. velj
    Odgovor korisniku/ci

    It’s always so amazing to me when folks continually argue for their limited beliefs. Think with a limited mindset and it becomes your truth, which also means we can flip that around and focus on abundance...

    1 proslijeđeni tweet 11 korisnika označava da im se sviđa
    Poništi
  4. proslijedio/la je Tweet
    4. velj

    Lmao.

    9 replies 370 proslijeđenih tweetova 1.055 korisnika označava da im se sviđa
    Poništi
  5. proslijedio/la je Tweet
    29. sij

    Remarkable. Electronic patient records systems used by thousands of doctors were programmed to automatically suggest opioids at treatment, thanks to a secret deal between the software maker and a drug company. reports. via

    167 replies 3.596 proslijeđenih tweetova 3.596 korisnika označava da im se sviđa
    Poništi
  6. proslijedio/la je Tweet
    15. kol 2019.

    "New State of the Art AI Optimizer: Rectified Adam (RAdam). Improve your AI accuracy instantly versus Adam, & why it works" It's been a long time since we've seen a new optimizer reliably beat the old favorites; this looks like a very encouraging approach!

    19 replies 919 proslijeđenih tweetova 2.606 korisnika označava da im se sviđa
    Poništi
  7. 31. sij

    Really enjoyed this episode. Workflow design and data pipeline automation of production workflows is a major challenge in production-scale machine learning projects. Great to see how Lyft is handling these challenges.

    Hey everyone, I'm so excited to share my recent interview on Scalable and Maintainable Workflows at Lyft with Flyte with for the podcast. Check it out! via
    1 reply
    Poništi
  8. 28. sij

    This is hilarious because of how accurate it is. I see opportunity here for a fork of Jupyter notebook that preserves ther original execution order of one's notebook. Immutable cell execution order... Wait, blockchain for Jupyter! Project name: ....Pepe Silvia

    Me explaining the execution order of my Jupyter notebook cells.
    Poništi
  9. proslijedio/la je Tweet
    26. sij

    . wrote a great post on the effect of data on improving deepfake performance. You can find his fork here: And visualize his results here:

    6 proslijeđenih tweetova 6 korisnika označava da im se sviđa
    Poništi
  10. proslijedio/la je Tweet
    25. sij

    You can become successful at something without knowing what you're doing. You can become successful at something without having much particular talent at it. But you can never become successful at anything without taking action. Ever.

    13 replies 534 proslijeđena tweeta 1.927 korisnika označava da im se sviđa
    Poništi
  11. proslijedio/la je Tweet
    24. sij

    I have thought about these two paragraphs so many times over the years. Huge thanks to for finding the source.

    31 reply 256 proslijeđenih tweetova 1.080 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  12. proslijedio/la je Tweet
    23. sij

    The January DVC Heartbeat is here. In this issue we're discussing , reaching 100 contributors, reducing technical debt, and learning about MLOps.

    9 proslijeđenih tweetova 20 korisnika označava da im se sviđa
    Poništi
  13. proslijedio/la je Tweet
    28. lip 2019.
    Odgovor korisniku/ci

    I think so, yes. VI when you mostly care about prediction (e.g. ) and sampling when you care about inference.

    2 proslijeđena tweeta 11 korisnika označava da im se sviđa
    Poništi
  14. proslijedio/la je Tweet
    16. lis 2019.

    Sage advice that maps easily to any Machine Learning product - start by deploying the simplest possible ‘toy model’ into production and then iterate

    4 proslijeđena tweeta 5 korisnika označava da im se sviđa
    Poništi
  15. 14. sij

    One trait I've observed important in software development is persistence and scrappiness. So much of working in tech is problem-solving. And not just the problem your project is trying to solve, but also all the bugs/inconsistencies/config issues that inevitably arise during it.

    Poništi
  16. proslijedio/la je Tweet
    13. sij

    I published a new article on the blog: Active Transfer Learning with PyTorch. Read about adapting Machine learning models with the knowledge that some data points will later get correct human labels, even if the model doesn't yet know the labels:

    1 reply 159 proslijeđenih tweetova 622 korisnika označavaju da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  17. proslijedio/la je Tweet
    14. sij

    Learn how to automate most of the infrastructure work required to deploy PyTorch models in production using Cortex, an open source tool for deploying models as APIs on AWS.

    90 proslijeđenih tweetova 334 korisnika označavaju da im se sviđa
    Poništi
  18. 14. sij

    3 — Usually outperform Adam in final results due to improved stability throughout training (via the longer term memory and clipping from it). Excited to implementing this new SOTA optimizer in future deep learning training.

    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  19. 14. sij

    AdaMod (Adam + memory) represents another step forward for deep learning optimizers as it provides three improvements: 1 — No need for warm-up (similar to Rectified Adam) 2 — Reduced sensitivity to learning rate hyperparam (converges to similar results)

    “Meet AdaMod: a new deep learning optimizer with memory” by Less Wright
    1 reply 1 korisnik označava da mu se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  20. proslijedio/la je Tweet
    22. pro 2019.
    Odgovor korisnicima

    Dear Cherry, Please retract the ACM endorsement of this letter. The ACM should be devoted to the open exchange of scientific information. The letter is an attempt to preserve the undue profiteering of commercial publishers.

    12 proslijeđenih tweetova 85 korisnika označava da im se sviđa
    Poništi
  21. proslijedio/la je Tweet
    12. sij

    Working with nbdev () is such a great experience! Finally breaking the cycle of starting to code in notebooks, then exporting the code to a library, and finally writing documentation. Awesome job & !

    29 proslijeđenih tweetova 183 korisnika označavaju da im se sviđa
    Poništi

@ChrisPoptic još nije objavio nijedan Tweet.

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:

    ·