Adaptive Path-Integral Autoencoders: Representation Learning and Planning for Dynamical Systems

Authors: Jung-Su Ha, Young-Jin Park, Hyeok-Joo Chae, Soon-Seo Park, Han-Lim Choi

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical experiments show that the proposed path-integral control based variational inference method leads to tighter lower bounds in statistical model learning of sequential data.
Researcher Affiliation Academia Jung-Su Ha, Young-Jin Park, Hyeok-Joo Chae, Soon-Seo Park, and Han-Lim Choi Department of Aerospace Engineering & KI for Robotics, KAIST Daejeon 305-701, Republic of Korea {{jsha, yjpark, hjchae, sspark}@lics., hanlimc@}kaist.ac.kr
Pseudocode Yes We implemented APIAE with Tensorflow [Abadi et al., 2016]; the pseudo code and algorithmic details of APIAE are given in the Appendix C.
Open Source Code Yes The supplementary video1 and the implementation code2 are available online. 2https://github.com/yjparkLiCS/18-NeurIPS-APIAE
Open Datasets Yes We utilized human motion capture data from the Carnegie Mellon University motion capture (CMU mocap) database for the learning;
Dataset Splits No The paper specifies training and test data amounts but does not mention a separate validation split for either experiment (e.g., Pendulum: '3000 and 500 data are used for training and test, respectively.' Mocap: '1043 and 173 data are used for training and test, respectively.').
Hardware Specification No The paper mentions that 'the whole operations were parallelized with GPU' but does not specify any particular GPU model, CPU, or other hardware components.
Software Dependencies No The paper mentions 'We implemented APIAE with Tensorflow [Abadi et al., 2016]' but does not provide a specific version number for TensorFlow or any other software.
Experiment Setup Yes We set our APIAE parameters as L=8, R=4, and K=10 during experiments.