Felix Yu

@flyyufelix

Founder of DisclosureTracker, a financial info tracking app for HK market. Applied Math at Stanford. Kaggle Master. Love to think about intelligence and ML

Vrijeme pridruživanja: listopad 2010.

Tweetovi

Blokirali ste korisnika/cu @flyyufelix

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

  1. Prikvačeni tweet
    11. lip 2019.

    A detailed account on the methods we tried to tackle overfitting and improve robustness for Donkey Car at Pixmoving Hackathon

    Poništi
  2. proslijedio/la je Tweet
    30. srp 2019.

    ML dev speed hack #1 - PyTorch over TF - Time to first step is faster b/c no static graph compilation - Easier to get loud errors via assertions within the code - Easier to drop into debugger and inspect tensors (TF2.0 may solve some of these problems but is still raw)

    Prikaži ovu nit
    Poništi
  3. proslijedio/la je Tweet
    24. srp 2019.
    Poništi
  4. 29. sij 2019.

    Great work! Use RL to train DonkeyCar that is applicable towards the real world. Can potentially train the car with RL in a matter of minutes on the real track, with smooth steering.

    Poništi
  5. proslijedio/la je Tweet
    13. lis 2018.

    This is how a mechanical binary counter works

    Poništi
  6. proslijedio/la je Tweet
    10. lis 2018.

    (1/N) S-RL Toolbox: Reinforcement Learning (RL) and State Representation Learning (SRL) Toolbox for Robotics Github: Doc: Paper: \w cc

    Prikaži ovu nit
    Poništi
  7. proslijedio/la je Tweet
    23. ruj 2018.
    Poništi
  8. 11. ruj 2018.

    I've hacked together a simulator to train Donkey car with Reinforcement Learning. I trained a Double Deep Q Learning (DDQN) agent to drive the car and it worked! The next step is to transfer the learned policy to the real world! Blog Post:

    Poništi
  9. proslijedio/la je Tweet
    5. srp 2018.

    . in case you guys enjoy challenging shadow/highlight patterns; we're on the verge of cracking that problem in France 🤓🇫🇷 Fun fact: this RC car was trained 100% in simulation and had never seen that shadow pattern before 💾

    Prikaži ovu nit
    Poništi
  10. proslijedio/la je Tweet
    26. lip 2018.

    I wrote a blog post on re-implementing 's Curiosity-Driven Exploration as training option in ML-Agents's PPO:

    Poništi
  11. proslijedio/la je Tweet
    22. lip 2018.
    Poništi
  12. 12. lip 2018.

    Blog post detailing the approach that got me a 5th place finish in OpenAI Retro Contest. We trained a reinforcement learning agent to play previously unseen custom levels of Sonic the Hedgehog using transfer learning.

    Poništi
  13. 18. stu 2017.

    I’ve implemented Directed Future Prediction (DFP), the RL model that won the ‘Full Deathmatch’ track of the VizDoom AI Competition in 2016. To better understanding DFP and my implementation in Keras, you can find my write-up here:

    Poništi
  14. 14. stu 2017.

    Very nice work! A must read for anyone who is interested in the subject

    Poništi
  15. proslijedio/la je Tweet
    29. lis 2017.
    Prikaži ovu nit
    Poništi
  16. 28. lis 2017.

    I’ve implemented DeepMind’s Distribution Bellman “C51” in Keras and tested on VizDoom

    Poništi
  17. 17. tra 2017.

    Detect and Classify Species of Fish from Fishing Vessels with Modern Object Detectors and ConvNets

    Poništi
  18. proslijedio/la je Tweet
    11. lis 2016.

    Deep Reinforcement Learning (DDPG) for 3D Car Racing Simulation TORCS with

    Poništi
  19. proslijedio/la je Tweet
    19. kol 2016.
    Poništi
  20. proslijedio/la je Tweet
    18. kol 2016.

    Just out: our new paper unlocks neural nets by removing the dependency on backpropagation using Synthetic Gradients!

    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:

    ·