GitPlanet

Cheap and reliable Node.js hosting starts at $3/month, and $1/month static HTML hosting

Created with love in Canada, visit hostnodejs.com today

Feel like to post an Ad? Learn Details
All Projects → nnaisense → Max

nnaisense / Max

Code for reproducing experiments in Model-Based Active Exploration, ICML 2019

Programming Languages

python
139335 projects - #7 most used programming language

Labels

reinforcement-learningdeep-reinforcement-learning

Projects that are alternatives of or similar to Max

Pytorch A3c
PyTorch implementation of Asynchronous Advantage Actor Critic (A3C) from "Asynchronous Methods for Deep Reinforcement Learning".
Stars: ✭ 879 (+1340.98%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Awesome Deep Rl
For deep RL and the future of AI.
Stars: ✭ 985 (+1514.75%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Drlkit
A High Level Python Deep Reinforcement Learning library. Great for beginners, prototyping and quickly comparing algorithms
Stars: ✭ 29 (-52.46%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Slm Lab
Modular Deep Reinforcement Learning framework in PyTorch. Companion library of the book "Foundations of Deep Reinforcement Learning".
Stars: ✭ 904 (+1381.97%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Deterministic Gail Pytorch
PyTorch implementation of Deterministic Generative Adversarial Imitation Learning (GAIL) for Off Policy learning
Stars: ✭ 44 (-27.87%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Paac.pytorch
Pytorch implementation of the PAAC algorithm presented in Efficient Parallel Methods for Deep Reinforcement Learning https://arxiv.org/abs/1705.04862
Stars: ✭ 22 (-63.93%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Rlcard
Reinforcement Learning / AI Bots in Card (Poker) Games - Blackjack, Leduc, Texas, DouDizhu, Mahjong, UNO.
Stars: ✭ 980 (+1506.56%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Deeprl Tutorials
Contains high quality implementations of Deep Reinforcement Learning algorithms written in PyTorch
Stars: ✭ 748 (+1126.23%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Deep Q Learning
Minimal Deep Q Learning (DQN & DDQN) implementations in Keras
Stars: ✭ 1,013 (+1560.66%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Deepbootcamp
Solved lab problems, slides and notes of the Deep Reinforcement Learning bootcamp 2017 held at UCBerkeley
Stars: ✭ 39 (-36.07%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Pygame Learning Environment
PyGame Learning Environment (PLE) -- Reinforcement Learning Environment in Python.
Stars: ✭ 828 (+1257.38%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Ml In Tf
Get started with Machine Learning in TensorFlow with a selection of good reads and implemented examples!
Stars: ✭ 45 (-26.23%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Btgym
Scalable, event-driven, deep-learning-friendly backtesting library
Stars: ✭ 765 (+1154.1%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Rl algos
Reinforcement Learning Algorithms
Stars: ✭ 14 (-77.05%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Osim Rl
Reinforcement learning environments with musculoskeletal models
Stars: ✭ 763 (+1150.82%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Left Shift
Using deep reinforcement learning to tackle the game 2048.
Stars: ✭ 35 (-42.62%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Pytorch Rl
Deep Reinforcement Learning with pytorch & visdom
Stars: ✭ 745 (+1121.31%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Tensorflow Tutorial
TensorFlow and Deep Learning Tutorials
Stars: ✭ 748 (+1126.23%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Deepqlearning.jl
Implementation of the Deep Q-learning algorithm to solve MDPs
Stars: ✭ 38 (-37.7%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
Async Deeprl
Playing Atari games with TensorFlow implementation of Asynchronous Deep Q-Learning
Stars: ✭ 44 (-27.87%)
Mutual labels:  reinforcement-learning, deep-reinforcement-learning
View All Similar Projects ➔

Model-Based Active Exploration (MAX)

Code for reproducing experiments in Model-Based Active Exploration, ICML 2019

Written in PyTorch v1.0.

Code relies on sacred for managing experiments and hyper-parameters.

Overview:

  • envs/: contains the environments used.
  • main.py: contains the main algorithm and baselines through modes.
  • models.py: a fast parallel implementation of an ensemble of models which can are trained with negative log-likelihood loss.
  • utilities.py: contains the all the utilities (exploration objectives) used in the paper.
  • imagination.py: contains code that constructs a virtual MDP using the model ensemble.
  • sac.py: contains a simple Soft Actor-Critic implementation.
  • sacred_fetcher.py: script to download experiment artifacts stored in MongoDB.

Installation

  • Install required dependencies:

    sudo apt install libosmesa6-dev libgl1-mesa-glx libglfw3 patchelf

  • Create conda environment with required dependencies:

    conda env create -f conda_env.yml

  • Download and setup MuJoCo binaries. The project uses mujoco and mujoco_py version 1.50.

    mkdir ~/.mujoco/
cd .mujoco/
wget -c https://www.roboti.us/download/mjpro150_linux.zip
unzip mjpro150_linux.zip
rm mjpro150_linux.zip

    Obtain MuJoCo license key and place it .mujoco/ directory created above with filename mjkey.txt.

  • Append the following to ~/.bashrc:

    # MuJoCo
export LD_LIBRARY_PATH=:/home/<USER>/.mujoco/mjpro150/bin

if [ -f /usr/lib/x86_64-linux-gnu/libGLEW.so ]; then    export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/home/<USER>/.mujoco/mjpro150/bin:/usr/lib/nvidia-390
    export LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libGLEW.so
    export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/lib/nvidia-375
fi

  • Quick test of MuJoCo installation

    >>> import gym
>>> gym.make('HalfCheetah-v2')

Commands

Execute the commands listed below from the code directory to reproduce the results.

Half Cheetah

  • MAX:
python main.py with max_explore env_noise_stdev=0.02
  • Trajectory Variance Active Exploration:
python main.py with max_explore utility_measure=traj_stdev policy_explore_alpha=0.2 env_noise_stdev=0.02
  • Renyi Divergence Reactive Exploration:
python main.py with max_explore exploration_mode=reactive env_noise_stdev=0.02
  • Prediction Error Reactive Exploration:
python main.py with max_explore exploration_mode=reactive utility_measure=pred_err policy_explore_alpha=0.2 env_noise_stdev=0.02
  • Random Exploration:
python main.py with random_explore env_noise_stdev=0.02

Ant

  • MAX:
python main.py with max_explore env_name=MagellanAnt-v2 env_noise_stdev=0.02 eval_freq=1500 checkpoint_frequency=1500 ant_coverage=True
  • Trajectory Variance Active Exploration:
python main.py with max_explore env_name=MagellanAnt-v2 utility_measure=traj_stdev policy_explore_alpha=0.2 env_noise_stdev=0.02 eval_freq=1500 checkpoint_frequency=1500 ant_coverage=True
  • Renyi Divergence Reactive Exploration:
python main.py with max_explore env_name=MagellanAnt-v2 exploration_mode=reactive env_noise_stdev=0.02 eval_freq=1500 checkpoint_frequency=1500 ant_coverage=True
  • Prediction Error Reactive Exploration:
python main.py with max_explore env_name=MagellanAnt-v2 exploration_mode=reactive utility_measure=pred_err policy_explore_alpha=0.2 env_noise_stdev=0.02 eval_freq=1500 checkpoint_frequency=1500 ant_coverage=True
  • Random Exploration:
python main.py with random_explore env_name=MagellanAnt-v2 env_noise_stdev=0.02 eval_freq=1500 checkpoint_frequency=1500 ant_coverage=True

Magellan

Magellan is the internal code name of the project inspired by life of Ferdinand Magellan.

Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].