Learning Temporally Causal Latent Processes from General Temporal Data

Authors: Weiran Yao, Yuewen Sun, Alex Ho, Changyin Sun, Kun Zhang

ICLR 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results on various datasets demonstrate that temporally causal latent processes are reliably identified from observed variables under different dependency structures and that our approach considerably outperforms baselines that do not properly leverage history or nonstationarity information.
Researcher Affiliation Academia Carnegie Mellon University, Pittsburgh PA, USA Southeast University, Nanjing, China Rice University, Houston TX, USA Mohamed bin Zayed University of Artificial Intelligence, Abu Dhabi, United Arab Emirates
Pseudocode No The paper describes its methods in prose and through diagrams (e.g., Figure 2 for LEAP architecture), but it does not include any structured pseudocode blocks or algorithms.
Open Source Code Yes Code: https://github.com/weirayao/leap
Open Datasets Yes Three public datasets including Ki TTi Mask (Klindt et al., 2020), Mass-Spring system (Li et al., 2020), and CMU Mo Cap database are used.
Dataset Splits Yes We use the ELBO loss on the validation dataset to select the best pair of [β, γ, σ] because low ELBO loss always leads to high MCC.
Hardware Specification Yes We used a machine with the following CPU specifications: Intel(R) Core(TM) i7-7700K CPU @ 4.20GHz; 8 CPUs, four physical cores per CPU, a total of 32 logical CPU units. The machine has two Ge Force GTX 1080 Ti GPUs with 11GB GPU memory.
Software Dependencies Yes The models were implemented in Py Torch 1.8.1.
Experiment Setup Yes The hyperparameters of LEAP include [β, γ, σ], which are the weights of each term in the augmented ELBO objective, as well as the latent size n and maximum time lag L. We use the ELBO loss on the validation dataset to select the best pair of [β, γ, σ]... A learning rate of 0.002 and a mini-batch size of 32 are used. For the noise discriminator, we use SGD optimizer with a learning rate of 0.001.