search

skumar9876 / Hierarchical-DQN

Implementation of the paper Hierarchical Deep Reinforcement Learning: Integrating Temporal Abstraction and Intrinsic Motivation - https://arxiv.org/pdf/1604.06057.pdf
79 stars 16 forks source link

clustering.py missing #1

Open THANNAGA opened 6 years ago

THANNAGA commented 6 years ago

Thank you for this code. Is the project still active, and does the code indeed reproduce the result of the paper? It looks like the clustering.py script is missing.

GoingMyWay commented 6 years ago

Indeed, clustering.py is missing

skumar9876 commented 6 years ago

I apologize for not seeing this earlier! Yes, clustering.py is missing - the Hierarchical DQN agent was used with subgoals formed by clustering the state space for an exploratory project. The purpose of this repo is to provide an implementation of a Hierarchical DQN agent, however this agent has not been tested with the experiments done in Kulkarni et al. The train_dqn file should be rewritten to remove the use of the clustering module.

On Mon, Jul 30, 2018 at 10:00 PM Alexander Yau notifications@github.com wrote:

Indeed, clustering.py is missing

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/skumar9876/Hierarchical-DQN/issues/1#issuecomment-409068885, or mute the thread https://github.com/notifications/unsubscribe-auth/AIMlfkVL2n8mZMOXQzwBkJpXPZjVjQuDks5uL7ojgaJpZM4UWplp .

YantianZha commented 5 years ago

@skumar9876 I agree with you. The original H-DQN uses the same state spaces for both meta-controller and low-level controller. So there is no need to use clustering.

I'm trying to use your code. So have you rewritten the code without the clustering module (i.e., the purely reimplementation of H-DQN)?

Thanks!

momofive commented 4 years ago

Excuse me, I wonder have you rewritten the code without the clustering module? Thanks!