Dropout Labs

@dropoutlabsai

We are building technology to train artificial intelligence on sensitive data without compromising privacy.

Halifax, Paris, San Francisco, and around the world
Vrijeme pridruživanja: listopad 2018.

Tweetovi

Blokirali ste korisnika/cu @dropoutlabsai

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

  1. Prikvačeni tweet
    10. sij

    2019 was a big year for privacy-preserving machine learning (PPML). Get caught up on the year’s developments with @jvmancuso’s year in review. THREAD 👇

    Prikaži ovu nit
    Poništi
  2. proslijedio/la je Tweet
    30. sij

    Thread: In this episode of I speak with , research scientist and creator of . We began by discussing how he found himself at the intersection of security and machine learning

    Prikaži ovu nit
    Poništi
  3. proslijedio/la je Tweet
    22. sij
    Odgovor korisnicima

    yes, is on 🔥🔥🔥!

    Poništi
  4. proslijedio/la je Tweet
    13. sij

    Halihax Presents: The Future of Application Development Tuesday, January 28 at 6:30pm Three greats talks on the subjects of React, GraphQL, AWS Fargate and Terraform Learn more about our speakers and RSVP to attend here:

    Prikaži ovu nit
    Poništi
  5. proslijedio/la je Tweet
    13. sij

    Watch this walkthrough of how to do collaborative fraud detection using TF Encrypted!

    Poništi
  6. 10. sij

    Here’s to another fast-paced year for PPML in 2020 🚀🎉🥳 With a focus on open science and further collaboration between researchers and practitioners, we’re looking forward to engineering security and privacy into the ML systems of today & tomorrow!

    Prikaži ovu nit
    Poništi
  7. 10. sij

    Tons of new open-source code! 😍 We got TF Federated and CrypTen, a variety of new packages within , an acceleration of grant-driven development within , and a host of new projects. Check the year-end review for more!

    Prikaži ovu nit
    Poništi
  8. 10. sij

    The gap between research and practice of PPML is as wide as ever, however applications of the tech beyond Google and Apple have shown up for the first time. We’re excited to see what new announcements 2020 will bring about for PPML adoption.

    Prikaži ovu nit
    Poništi
  9. 10. sij

    One new trend in research we’re particularly excited about is the development of new ML techniques that are privacy-oriented. If you’re a researcher in this line of work, we encourage you to submit to the ICLR Trustworthy ML Workshop by Jan 31!

    Prikaži ovu nit
    Poništi
  10. 10. sij

    Exciting progress has been made: encrypted inference of modern neural networks has become practical using both multi-party computation and secure enclaves, and new variants of differential privacy aim to make privacy practical.

    Prikaži ovu nit
    Poništi
  11. 10. sij

    Privacy was once again a major theme of mainstream media coverage of tech. The Great Hack made waves after its Netflix release, released a monumental work on Surveillance Capitalism, and Facebook announced a new privacy-first vision for social media.

    Prikaži ovu nit
    Poništi
  12. 23. pro 2019.

    We’re extremely excited for you to join us ! Checkout her announcement post below 👇

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

    Great framework on privacy preserving machine learning, wouldn’t it be great if we could live in a default encrypted world where models could run and train on encrypted data, that’s why I’m excited about

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

    We are super excited to introduce the Alibaba Gemini Lab from as our newest organizational contributor! They recently obtained 1st place in the iDASH competition (more to come!) and are now generously contributing the extension of used! 🙌😍

    Poništi
  15. 18. pro 2019.

    In the future, we hope to have first-class support for federated learning in TF Encrypted by fully integrating with TF Federated. We have an RFC with a suggested design - we’d love your feedback!

    Prikaži ovu nit
    Poništi
  16. 18. pro 2019.

    We've implemented a secure, federated version of Reptile to demonstrate this extensibility. We encourage you to write your own and consider sharing them with the community in Slack or as a PR!

    Prikaži ovu nit
    Poništi
  17. 18. pro 2019.

    We've written our example code with extensibility in mind, so that users can easily experiment with their own custom federated algorithms.

    Prikaži ovu nit
    Poništi
  18. 18. pro 2019.

    In normal FL, each client must implicitly trust a central party not to reconstruct its data from its model update. With secure aggregation, we can remove this trust. We use a secret sharing protocol from to perform the secure aggregation.

    Prikaži ovu nit
    Poništi
  19. 18. pro 2019.

    We’re publishing a blog post today about bringing federated learning (FL) and secure aggregation together in 2.0. It contains a detailed explanation of both concepts and a full example.

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

    great thread from @jvmancuso from on the latest at NeurIPS conference

    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:

    ·