ODE-RSSM: Learning Stochastic Recurrent State Space Model from Irregularly Sampled Data

Authors: Zhaolin Yuan, Xiaojuan Ban, Zixuan Zhang, Xiaorui Li, Hong-Ning Dai

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We also conduct extensive experiments to evaluate the proposed ODE-RSSM and the baselines on three input-output datasets, one of which is a rollout of a private industrial dataset with strong long-term delay and stochasticity. The results demonstrate that the ODE-RSSM achieves better performance than other baselines in open loop prediction even if the time spans of predicted points are uneven and the distribution of length is changeable.
Researcher Affiliation Academia 1School of Intelligence Science and Technology, Beijing Key Laboratory of Knowledge Engineering for Materials Science, University of Science and Technology Beijing, Beijing 100083, China. 2Beijing Advanced Innovation Center for Materials Genome Engineering, University of Science and Technology Beijing. 3Key Laboratory of Intelligent Bionic Unmanned Systems, Ministry of Education, University of Science and Technology Beijing, Beijing 100083, China 4Department of Computer Science, Hong Kong Baptist University, Hong Kong, China
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code Yes Code is availiable at https://github.com/yuanzhaolin/ODE-RSSM.
Open Datasets Yes We use three input/output datasets for conducting the experiments. Two of them, CSTR (one in two out) and Winding (five in two out), are public datasets (Demeester 2020).
Dataset Splits Yes Each dataset is split to three partitions for training (the foremost 60%), validation (the middle 20%), and test (the last 20%).
Hardware Specification No The paper mentions 'GPU memory' and 'CUDA parallelization' but does not provide specific hardware details such as exact GPU/CPU models, processor types, or memory amounts used for running experiments.
Software Dependencies No The paper mentions the use of 'Adam optimizer' but does not specify any software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup Yes Each dataset is split to three partitions for training (the foremost 60%), validation (the middle 20%), and test (the last 20%). The basic time difference |ti+1 ti| between any adjacent sampling points is uniformly defined as 0.1 for all datasets. All of the models are trained by the Adam optimizer where the learning rate is 5e-4. The training does not stop until the validation loss increases for 100 epochs. For the ODE-RSSM and the discrete-time models with stochastic states transitions, we repeatedly predict ntraj = 32 trajectories in parallel from the generative model and measure the mean prediction error between the sampled trajectories and the ground-truth.