Summary Markov Models for Event Sequences

Authors: Debarun Bhattacharjya, Saurabh Sihag, Oktie Hassanzadeh, Liza Bialik

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

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct an experimental investigation comparing the proposed models with relevant baselines, and illustrate their knowledge acquisition and discovery capabilities through case studies involving sequences from text. ... 5 Experiments The following experiments assess our two proposed Su MMs as well as our learning approach.
Researcher Affiliation Collaboration Debarun Bhattacharjya1 , Saurabh Sihag2 , Oktie Hassanzadeh1 and Liza Bialik3 1IBM Research 2University of Pennsylvania 3University of Massachusetts, Amherst
Pseudocode Yes Algorithm 1 Greedy Score-based Search and procedure COMPUTESCORE are presented in Section 4.
Open Source Code No The paper does not provide an explicit statement about the release of source code or a direct link to a code repository for the methodology described.
Open Datasets Yes Datasets. We consider the following structured datasets, some derived from time-stamped event datasets where the time stamp is ignored (assumed to be missing or erroneous). Diabetes [Frank and Asuncion, 2010]: Events for diabetic patients around meal ingestion, exercise, insulin intake and blood glucose level transitions b/w low, medium and high. Linked In [Xu et al., 2017]: Employment related events such as joining a new role for 1000 Linked In users. Stack Overflow [Grant and Betts, 2013]: Events for engagement of 1000 users (chosen from [Du et al., 2016]) around receipt of badges in a question answering website.
Dataset Splits Yes For experiments, each dataset is split into train/dev/test sets (70/15/15)%, removing labels that are not common across all three splits.
Hardware Specification No The paper does not explicitly describe the hardware used for running its experiments, such as specific CPU or GPU models, or memory specifications.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., library names with versions) needed to replicate the experiment.
Experiment Setup Yes For experiments, we use a single parameter α as hyper-parameter, assuming that αx|s = α x, s. ... BIC score penalizes models that are overly complex... where γ is a penalty on complexity, generally set at a default value of 1 unless otherwise specified.