Cardinality-Regularized Hawkes-Granger Model

Authors: Tsuyoshi Ide, Georgios Kollias, Dzung Phan, Naoki Abe

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We validate the proposed framework with two real use-cases, one from the power grid and the other from the cloud data center management domain.
Researcher Affiliation Industry Tsuyoshi Idé IBM Research, T. J. Watson Research Center tide@us.ibm.com Georgios Kollias IBM Research, T. J. Watson Research Center gkollias@us.ibm.com Dzung T. Phan IBM Research, T. J. Watson Research Center phandu@us.ibm.com Naoki Abe IBM Research, T. J. Watson Research Center nabe@us.ibm.com
Pseudocode Yes Algorithm summary Algorithm 1 summarizes L0Hawkes, the proposed algorithm, which is used as part of the iterative MM procedure in Eq. (13).
Open Source Code No The paper does not provide a direct link to a code repository for the proposed method nor explicitly state that the code is publicly available.
Open Datasets Yes We generated two synthetic multivariate event datasets, Sparse5 and Dense10, with a standard point process simulator tick [4]. We obtained failure event data (Grid) of power grid from U.S. Department of Energy [29].
Dataset Splits No The paper discusses cross-validation for parameter determination ('These parameters should eventually be cross-validated with independent episodes of event data, or ground through causality data.') but does not provide specific percentages or counts for training, validation, and test splits for the datasets used in experiments.
Hardware Specification Yes The mean computational time was (46, 881, 382) seconds per one parameter set for (L0Hawkes, c MLP, c LSTM), respectively, on a laptop PC (i7 CPU, 32GB memory, Quadro P3200 GPU).
Software Dependencies No The paper mentions using 'tick' [4], which is a Python library, but does not specify a version number for it or for any other software dependencies crucial for reproducibility.
Experiment Setup Yes We chose νA, νβ, νµ to be 0.1 and tested τ = 0.5, 1, 2. We grid-searched the model parameters based on AIC to get 5 (10−3, 10−4, 10−4) for (νµ, νβ, µA) and (1, 1) for (τ, ϵ).