DescriptionMy research focuses on detecting patterns in physiological data of patients in an Intensive Care Unit setting, with the aim of constructing an early warning system. The approach I am taking includes machine learning algorithms such as Artificial Neural Networks, Hidden Markov Models, and Support Vector Machines. I presently have limited computational resources for which to conduct this research. The data I am using for training and validation is both static and time based information on 32,000 patients and includes approximately 30GB of raw data. Additionally I have extremely high resolution data on 2,600 patient. The search-space is prohibitively large for a single computer and even some of the smaller clusters. I am employing an optimization methodology which allows for a differential evolution approach to incrementally improve a structurally adaptive model. The methodology allows for parallel programming which is of course a necessity for distributed computing.
OrganizationUniversity of Washington
DepartmentBiomedical and Health Informatics
Sponsor Campus GridOSG Connect
Principal Investigator
Karl Jablonowski
Field Of ScienceBioinformatics