Learning Continuous Control Policies by Stochastic Value Gradients

Authors: Nicolas Heess, Gregory Wayne, David Silver, Timothy Lillicrap, Tom Erez, Yuval Tassa

NeurIPS 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We apply these algorithms first to a toy stochastic control problem and then to several physics-based control problems in simulation. One of these variants, SVG(1), shows the effectiveness of learning models, value functions, and policies simultaneously in continuous domains.
Researcher Affiliation Industry Google Deep Mind {heess, gregwayne, davidsilver, countzero, tassa, etom}@google.com
Pseudocode Yes Algorithm 1 SVG(1), Algorithm 2 SVG(1) with Replay
Open Source Code No The paper provides a link to a video montage ('https://youtu.be/PYd L7bcn_c M.') but no explicit link or statement about open-sourcing the code used for the described methodology.
Open Datasets No The paper refers to using environments from the 'Mu Jo Co simulator' for experiments but does not provide access information (link, DOI, citation) for specific datasets used for training.
Dataset Splits No The paper does not provide specific percentages or counts for training, validation, or test data splits. It describes continuous interaction with simulation environments.
Hardware Specification No The paper does not provide any specific details about the hardware used to run the experiments, such as CPU/GPU models, memory, or cloud resources.
Software Dependencies No The paper mentions the 'Mu Jo Co simulator' and 'neural networks' but does not specify any software names with version numbers for reproducibility (e.g., Python 3.x, PyTorch 1.x).
Experiment Setup Yes In all cases, we use generic, 2 hidden-layer neural networks with tanh activation functions to represent models, value functions, and policies. With simulation time step 0.01s