Fritz Obermeyer

@ftzo

Vrijeme pridruživanja: siječanj 2011.
5 fotografija ili videozapisa Fotografije i videozapisi

Tweetovi

Blokirali ste korisnika/cu @ftzo

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

  1. Prikvačeni tweet
    10. kol 2017.

    every software project ever

    1 reply 3 korisnika označavaju da im se sviđa
    Poništi
  2. 17. sij

    Pyro 1.2 release adds poutine.reparam to rewrite models to improve geometry via: - neural transport - discrete cosine transform - auxiliary variable methods for Levy Stable distributions - conditional Gaussian HMMs - decentering - transform unwrapping

    42 proslijeđena tweeta 184 korisnika označavaju da im se sviđa
    Poništi
  3. proslijedio/la je Tweet
    5. sij

    I running a 60-minute webinar blitz on causal modeling in machine learning code examples THIS Thursday. Ideal for applied and practitioners interested in connections between dense causal inference lit and ML tools practice.

    5 replies 20 proslijeđenih tweetova 51 korisnik označava da mu se sviđa
    Poništi
  4. 16. pro 2019.

    This is a great idea! We just added it to Pyro as a poutine.reparam effect. First use case was an auxiliary variable reparameterization for the Levy stable distribution, where .log_prob() is not analytic.

    Want to automatically transform your probabilistic programs for more efficient inference? Check out our new paper on automatic reparameterisation: ; with and Matt Hoffman.
    Prikaži ovu nit
    2 proslijeđena tweeta 20 korisnika označava da im se sviđa
    Poništi
  5. proslijedio/la je Tweet
    7. pro 2019.

    Reweighted wake-sleep is now in Pyro :)

    Pyro 1.1.0 is released: three new inference algorithms, a pyro.deterministic() primitive, new heavy-tailed distributions, Causal Effect VAEs, and easier ways to make PyroModules. Thanks to Siddharth Narayanaswamy, Tuan Anh Le, and other contributors!
    10 proslijeđenih tweetova 52 korisnika označavaju da im se sviđa
    Poništi
  6. 6. pro 2019.

    Pyro 1.1.0 is released: three new inference algorithms, a pyro.deterministic() primitive, new heavy-tailed distributions, Causal Effect VAEs, and easier ways to make PyroModules. Thanks to Siddharth Narayanaswamy, Tuan Anh Le, and other contributors!

    28 proslijeđenih tweetova 130 korisnika označava da im se sviđa
    Poništi
  7. 15. stu 2019.

    Pyro 1.0 is released: stable APIs, a jit-compatible Predictive helper, new normalizing flows, parallel-scan GPs and state space models, more-automatic AutoGuides, an OED tutorial, FoldedDistribution, and PyroModule[-] to make your nn.Modules Bayesian

    1 reply 97 proslijeđenih tweetova 385 korisnika označava da im se sviđa
    Poništi
  8. proslijedio/la je Tweet
    28. lis 2019.

    Just having fun using NumPyro and JAX to do inference for ODE parameters - this finishes the translation of 's Statistical Rethinking 2nd ed. book (latest draft version) to NumPyro

    30 proslijeđenih tweetova 123 korisnika označavaju da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  9. 24. lis 2019.

    We'd love feedback on the initial release of Funsor, a tensor-like library for functions and distributions. Possibly "Distributions 2.0". We use it with Pyro, but it's also fun for stand-alone Gaussian computations. code: paper:

    1 reply 41 proslijeđeni tweet 177 korisnika označava da im se sviđa
    Poništi
  10. 23. lis 2019.

    Pyro 0.5 is released: pyro.factor statement, conditional MADE autoregressive nn, OED tutorial, kl_divergence for Independent & Delta, CRPS to calibrate sample sets, moved pyro.generic to a new package pyro-api, reorganized pyro.distributions

    28 proslijeđenih tweetova 126 korisnika označava da im se sviđa
    Poništi
  11. 17. lis 2019.

    How can I determine who the clever lawyer was behind Zappos' class action case, which concluded in Zappos being forced to offer a limited-time 10% discount coupon to customers whose data had been leaked —a gov-endorsed coupon that bypasses spam filters?

    1 reply 2 korisnika označavaju da im se sviđa
    Poništi
  12. proslijedio/la je Tweet
    11. ruj 2019.

    Utilizing JAX pmap, map, vmap, we can draw chains in sequential, or parallel with many devices (CPUs/GPUs), or even with a single device using vectorization. This cool feature is available in NumPyro 0.2.0. See the release note

    1 reply 7 proslijeđenih tweetova 32 korisnika označavaju da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  13. 15. kol 2019.

    Course material is at (latest draft, feedback and PRs welcome 🙂)

    1 proslijeđeni tweet 7 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  14. 15. kol 2019.

    Want to learn about Time Series Forecasting in Pyro? Join and I for a tutorial in San Diego next Tuesday 8/20

    1 reply 8 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  15. proslijedio/la je Tweet
    7. kol 2019.

    Join us at Machine Learning Summit hosted by Uber in San Diego August 20th. Open Source talks about , , and , hands-on workshops on and . Uber will donate your registration fee to !

    11 proslijeđenih tweetova 9 korisnika označava da im se sviđa
    Poništi
  16. 18. srp 2019.

    I'm loving the time-parallel inference based on Särkkä & García-Fernández This makes it super cheap to do Baum-Welch training of neural HMMs on GPUs. Try it out with pyro.distributions.DiscreteHMM

    7 korisnika označava da im se sviđa
    Poništi
  17. 16. srp 2019.

    Pyro 0.3.4 is released! A flexible easyguide module, customizable autoguide initialization, more normalizing flows, new schedulers, faster HMM learning, and a redesigned MCMC interface. This is our last Python 2 compatible release.

    57 proslijeđenih tweetova 230 korisnika označava da im se sviđa
    Poništi
  18. 3. lip 2019.

    The Pyro team is excited to release NumPyro, a JAX backend for the Pyro probabilistic programming language. The modeling language mirrors Pyro, and inference via HMC and NUTS is very fast. Let us know what you think!

    86 proslijeđenih tweetova 301 korisnik označava da mu se sviđa
    Poništi
  19. 24. tra 2019.

    Pyro 0.3.2 is released! Capture-recapture example, new normalizing flows, faster and easier discrete inference, LKJ and SpanningTree distributions, diagnostics for VI, and a pyro.generic interface to support new backends for Pyro

    14 proslijeđenih tweetova 66 korisnika označava da im se sviđa
    Poništi
  20. proslijedio/la je Tweet
    7. tra 2019.

    So excited is becoming more and more popular. Now caught up by team in . For what I value is their core idea of API replication. This is a more than a decade of scientific computing evolution, follow! Thanks for sharing

    16 proslijeđenih tweetova 68 korisnika označava da im se sviđa
    Poništi
  21. proslijedio/la je Tweet
    13. velj 2019.

    New work with the Pyro team () on efficient batched discrete variable inference through tensor variable elimination. Practically generalizes 'einsum' with batching, semirings (log-space, max, etc), and marginals. Powers Pyro's discrete directed inference.

    27 proslijeđenih tweetova 94 korisnika označavaju da im se sviđa
    Poništi

@ftzo 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:

    ·