Giovanni Neglia

@GiovanniNeglia1

Researcher at Sophia-Antipolis, France

Alpes-Maritimes
Vrijeme pridruživanja: listopad 2018.

Tweetovi

Blokirali ste korisnika/cu @GiovanniNeglia1

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @GiovanniNeglia1

  1. proslijedio/la je Tweet
    1. velj

    Will be at ITA next week to present our recent paper on Federated Learning (). This work is accepted to AISTATS 2020.

    Poništi
  2. proslijedio/la je Tweet
    1. velj

    “Come sempre succede in caso di calamità, ognuno pensava che se fosse dovuta arrivare una disgrazia avrebbe colpito il vicino, non lui” e la storia di tre generazioni di una famiglia ebraica tra il 1860 ed il 1940.

    Poništi
  3. proslijedio/la je Tweet
    26. sij

    I was invited by to give a talk at EISTI about Survival Analysis and Machine Learning. Here are my slides:

    Poništi
  4. 13. sij

    Presque une heure après l’accident signale que aucun bus circulera suite au « droit de retrait du personnel de conduite ». Pas clair parce que toutes les lignes devraient s’arrêter...

    Prikaži ovu nit
    Poništi
  5. 13. sij

    À tous les personnes qui utilisent la ligne A pour aller à Sophia Antipolis. Il y a eu un accident sur la voie réservée aux bus avant d'arriver à Saint Philippe. Les bus ne passent pas dans les 2 directions.

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

    Journée « Mathématiques et Intelligence Artificielle » le lundi 9 mars au Muséum d’Histoire Naturelle à Paris :

    Poništi
  7. proslijedio/la je Tweet
    11. sij

    My article "Narrating Trauma" has just been published in Consecutio Rerum

    Poništi
  8. proslijedio/la je Tweet
    9. sij

    Because there is nothing more useful than a mathematical trick, and thanks to your feedback, find the updated ML trick slide -:) Please RT, add even more to the list !!!

    Poništi
  9. proslijedio/la je Tweet
    18. pro 2019.

    Une nouvelle version de mon graphique sur l'évolut° du salaire des enseignants depuis 1983, comparé à l'évolut°du salaire moyen. J'ai corrigé des données erronées pour 1988 et 1990 (j'avais compté 2 fois les primes) et j'ai précisé la légende. Commentaires à venir en dessous 🔽

    Prikaži ovu nit
    Poništi
  10. proslijedio/la je Tweet
    17. pro 2019.

    Maureen Clerc opening the talk at

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

    (Google) and yours truly show how to get variance reduction in SGD *without* inner loops, thanks to momentum! We get optimal* rates for stochastic nonconvex optimization. Poster #214, 5:00 -- 7:00 PM * Optimality just proven in

    Prikaži ovu nit
    Poništi
  12. proslijedio/la je Tweet
    9. pro 2019.

    There are now 3 papers that successfully use Self-Supervised Learning for visual feature learning: MoCo: PIRL: And this, below. All three use some form of Siamese net.

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

    In case you need to teach distributed DL training, here's a quick summary of different synchronization strategies (tree and ring sync, NVLink, Horovod, etc.).

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

    Stein's Lemma: If X ~ N(a, s^2), then s^2 * E[g'(X)] = E[g(X) * (X-a)] Corollary: Simple procedure for obtaining gradient-free unbiased estimator for expected gradient of function of Gaussian r.v.

    Poništi
  15. proslijedio/la je Tweet
    5. pro 2019.

    Meet the 'double descent' phenomenon. After we figure it out we should probably rewrite the book chapter on bias-variance tradeoff.

    Poništi
  16. proslijedio/la je Tweet
    5. pro 2019.

    1/ Why do wide, random neural networks form Gaussian processes, *regardless of architecture*? Let me give an overview in case you are too lazy to check out the paper or the code . The proof has two parts…

    Prikaži ovu nit
    Poništi
  17. proslijedio/la je Tweet
    1. pro 2019.
    Poništi
  18. proslijedio/la je Tweet
    1. pro 2019.

    Lost Connections de m'a laissée perplexe, voici pourquoi :

    Poništi
  19. proslijedio/la je Tweet

    Our browser extension for Firefox and Chrome (hopefully, coming soon on add-on stores) called "Cookie Glasses" allows users to verify that the consent stored by CMPs corresponds to their choice: (10/11)

    Prikaži ovu nit
    Poništi
  20. proslijedio/la je Tweet
    27. stu 2019.
    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:

    ·