Addressing Sample Complexity in Visual Tasks Using HER and Hallucinatory GANs

Authors: Himanshu Sahni, Toby Buckley, Pieter Abbeel, Ilya Kuzovkin

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We validate our approach on 3D navigation tasks and a simulated robotics application and show marked improvement over baselines derived from previous work. 6 Experiments We test our method on two first person visual environments. In a modified version of Mini World [6], we design two tasks.
Researcher Affiliation Collaboration Himanshu Sahni Toby Buckley Pieter Abbeel , Ilya Kuzovkin Georgia Institute of Technology Off World Inc. University of California, Berkeley Work done as an intern at Off World Inc. Correspondence to: hsahni3@gatech.edu
Pseudocode Yes Algorithm 1 HALGAN+HER
Open Source Code No The paper does not provide a direct link or explicit statement about the availability of its source code.
Open Datasets No HALGAN is trained on a dataset, R, of observations of the goal where the relative configuration to the agent is known. For the purposes of our experiments, we collect the training data in R by using the last 16 or 32 states of a successful rollout. The paper describes generating its own dataset within open environments but does not provide access to the collected dataset itself.
Dataset Splits No The paper describes data collection and training processes but does not specify explicit train/validation/test dataset splits with percentages, counts, or references to predefined splits.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU models, CPU types, memory) used for running the experiments.
Software Dependencies No The paper mentions software like 'Mini World', 'Gazebo', 'gym-gazebo', 'DDQN', and 'DDPG' but does not provide specific version numbers for these or any other software dependencies.
Experiment Setup No Details of all experimental hyperparameters are provided in the appendix. This indicates that specific setup details are not present in the main text of the paper.