Tower of Bleyddyn ap Rhys

  • Projects
  • CV
  • Blog
  • ML Notes
  1. On the link between conscious function and general intelligence in humans and machines

    Published: Fri 01 July 2022
    By Bleyddyn

    Paper

    tags: ML Papers

    Access consciousness and its relation to general intelligence

    read more
  2. 'Fastai Transforms for DonkeyCar'

    Published: Sun 13 March 2022
    By Bleyddyn

    tags: malpi fastai augmentation ML

    Fastai transforms don't seem to help DonkeyCar training

    read more
  3. 'Effect of Dropout Layers in a VAE'

    Published: Tue 14 January 2020
    By Bleyddyn

    tags: malpi vae ML

    Dropout layers make VAE output worse

    read more
  4. Stable Baselines Algorithms

    Published: Sun 03 February 2019
    By Bleyddyn

    tags: baselines RL ML

    Table of algorithms in the Stable Baselines repo

    read more
  5. Unicorn: Continual learning with a universal, off-policy agent

    Published: Thu 19 July 2018
    By Bleyddyn

    Paper

    tags: ML Papers RL

    Continual learning with a universal, off-policy agent.

    read more
  6. Sample-Efficient Deep RL with Generative Adversarial Tree Search

    Published: Wed 27 June 2018
    By Bleyddyn

    Paper

    tags: ML Papers RL

    Learned dynamics model with a GAN for image generation and MCTS for planning.

    read more
  7. Learning Real-World Robot Policies by Dreaming

    Published: Mon 25 June 2018
    By Bleyddyn

    Paper

    tags: ML Papers

    Unsupervised learning of image encoding, dynamics and reward models.

    read more
  8. Unsupervised Predictive Memory in a Goal-Directed Agent

    Published: Thu 12 April 2018
    By Bleyddyn

    Paper

    tags: ML Papers

    Unsupervised training of a memory that is used for prediction of state and reward.

    read more
  9. World Models

    Published: Mon 09 April 2018
    By Bleyddyn

    Paper

    tags: ML Papers RL

    Unsupervised learning of image encoding and dynamics model.

    read more
  10. 'First Real DAgger Results'

    Published: Mon 12 March 2018
    By Bleyddyn

    tags: malpi dagger ML

    Using DAgger to improve MaLPi's training, while MaLPi is driving

    read more
  11. 'Initial DAgger Results'

    Published: Tue 27 February 2018
    By Bleyddyn

    tags: malpi dagger ML

    Using DAgger to improve MaLPi's training

    read more
  12. 'First attempts at Hyperparameter Optimization'

    Published: Sat 03 February 2018
    By Bleyddyn

    tags: numpy malpi hyperparameters hyperopt ML

    Hyperparameter optimization using hyperopt on racetrack data.

    read more
  13. Pretraining Deep Actor-Critic Reinforcement Learning Algorithms With Expert Demonstrations

    Published: Thu 01 February 2018
    By Bleyddyn

    Paper

    tags: ML Papers RL

    Pretraining Deep Actor-Critic Reinforcement Learning Algorithms With Expert Demonstrations.

    read more
  14. 'Normalizing image data before training - GRU version'

    Published: Sun 24 December 2017
    By Bleyddyn

    tags: numpy malpi gru ML

    Testing image normalization when using a GRU.

    read more
  15. 'Normalizing image data before training - LSTM version'

    Published: Thu 21 December 2017
    By Bleyddyn

    tags: numpy malpi lstm ML

    Testing image normalization when using an LSTM.

    read more
  16. Tensorizing LSTMs

    Published: Mon 18 December 2017
    By Bleyddyn

    Paper

    tags: ML Papers RNN

    Tensorizing LSTMs to make them wider and deeper without adding parameters and with minimal extra compute costs.

    read more
  17. 'Normalizing image data before training'

    Published: Wed 13 December 2017
    By Bleyddyn

    tags: numpy malpi ML

    I had completely forgotten to normalize the images I'm feeding into MaLPi's network, so I thought I'd try to be a bit more formal about it than my usual.

    read more
  18. Mastering the game of Go without human knowledge

    Published: Thu 26 October 2017
    By Bleyddyn

    Paper

    tags: ML Papers RL

    AlphaGo Zero, all RL self-play.

    read more
  19. Getting a Keras LSTM layer to work on MaLPi

    Published: Thu 19 October 2017
    By Bleyddyn

    tags: keras lstm malpi ML

    Training on batch sizes and/or sequence lengths longer than one, while still being able to run one image at a time on the robot.

    read more
  20. Pre-training Neural Networks with Human Demonstrations for Deep Reinforcement Learning

    Published: Thu 14 September 2017
    By Bleyddyn

    Paper

    tags: ML Papers RL

    Pre-train using supervised learning on human provided demonstations.

    read more

Page 1 / 2 | Next | Last

Bleyddyn

Bleyddyn

Professional Programmer, ML Student

social

  • Site Feed (Atom)
  • Mastodon
  • Codeberg
  • Github

links

  • Pelican
  • Python.org
  • Theme by Smashing Magazine