A Multitask Point Process Predictive Model

Authors: Wenzhao Lian, Ricardo Henao, Vinayak Rao, Joseph Lucas, Lawrence Carin

ICML 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results are shown on both synthetic data and as well as real electronic health-records data. We evaluate our model on both synthetic data as well as an EHR dataset.
Researcher Affiliation Academia 1Department of Electrical and Computer Engineering, Duke University, Durham, NC 27708, USA 2Department of Statistics, Purdue University, West Lafayette, IN 47907, USA 3Center for Predictive Medicine, Duke Clinical Research Institute, Durham, NC 27708, USA
Pseudocode No The paper describes the variational EM algorithm steps in prose within Section 5 "Inference" and provides a graphical model in Figure 1, but it does not include a formally structured pseudocode block or algorithm listing.
Open Source Code No The paper does not contain any explicit statement about releasing source code for the methodology described, nor does it provide a link to a code repository.
Open Datasets Yes We use the New Zealand national minimum dataset 1, covering the years 2007 through 2011 (inclusive). The data contains approximately 3.3 million inpatient visits from 1.5 million unique individuals with ages from 18 to 65. Available variables include ICD-10-AM (Australian Modification) diagnosis and procedure codes which are grouped into 22 broad categories (World Health Organization, 2010). Footnote 1: http://www.health.govt.nz/nz-health-statistics
Dataset Splits Yes split visit sequences for each patient into training and testing (split at the time stamp when half of the number of visits are observed).
Hardware Specification No The paper does not provide specific details about the hardware used to run the experiments, such as GPU models, CPU types, or memory specifications. It only describes the algorithms and datasets.
Software Dependencies No The paper does not specify any software dependencies with version numbers (e.g., Python 3.8, TensorFlow 2.x, etc.) that would be necessary for reproducibility.
Experiment Setup Yes For inference, the pseudo points were initialized using Kmeans clustering over the history features appearing in the training set. The GP hyperparameters are initialized setting the length-scale parameters, {λp}, as the standard deviation of observed feature vectors, the magnitude, τ, as the standard deviation of square roots of empirical rates (computed via binning methods), and the noise term, σ2, as a small value, 0.01τ 2. For inference, we set the number of pseudo inputs M = 15.