Modeling Correlated Arrival Events with Latent Semi-Markov Processes

Authors: Wenzhao Lian, Vinayak Rao, Brian Eriksson, Lawrence Carin

ICML 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental apply our ideas to both synthetic data and a real-world biometrics application. We evaluate the performance of our model and inference methodology on both synthetic and real-world biometrics data.
Researcher Affiliation Collaboration Wenzhao Lian WL89@DUKE.EDU Department of Electrical and Computer Engineering, Duke University, Durham, NC 27708, USA Vinayak Rao VAR11@STAT.DUKE.EDU Department of Statistical Science, Duke University, Durham, NC 27708, USA Brian Eriksson BRIAN.ERIKSSON@TECHNICOLOR.COM Technicolor Research Center, 735 Emerson Street, Palo Alto, CA 94301, USA Lawrence Carin LCARIN@DUKE.EDU Department of Electrical and Computer Engineering, Duke University, Durham, NC 27708, USA
Pseudocode Yes Algorithm 1 in the appendix gives details of this generative process.
Open Source Code Yes Code available at http://people.duke.edu/~wl89/
Open Datasets No The paper uses synthetic data and a custom-collected biometrics dataset but does not provide concrete access information or state its public availability.
Dataset Splits No The paper describes how the synthetic data was generated and the MCMC iterations (e.g., 'discarding the first 2000 as burn-in'), but does not specify explicit training, validation, and test splits for the real-world dataset.
Hardware Specification Yes The running time of a typical trial (with T = 1000 and about 120 event arrivals for each user) was about 3000 seconds with unoptimized Matlab code on a computer with 2.2GHz CPU and 8GB RAM.
Software Dependencies No The paper mentions 'unoptimized Matlab code' but does not specify the version number of Matlab or any other software dependencies with their versions.
Experiment Setup Yes For inference, the fixed hyperparameters of the sampler were set as: α = 3, c = d = e = f = 10 3, and πk = [0.5, 0.5]T . We ran 5000 MCMC iterations of our MCMC sampler, discarding the first 2000 as burn-in, with posterior samples collected every 5 iterations.