2018/01/23 by L. Mihaela Păun, Paun, L. Mihaela, M. Umar Qureshi +11 · 1 citation
Computer Science · Decision Sciences · #Machine Learning in Healthcare #demographic modeling and climate adaptation
paper · pdf · doi:10.48550/arxiv.1801.07742
This study performs parameter inference in a partial differential equations\nsystem of pulmonary circulation. We use a fluid dynamics network model that\ntakes selected parameter values and mimics the behaviour of the pulmonary\nhaemodynamics under normal physiological and pathological conditions. This is\nof medical interest as it enables tracking the progression of pulmonary\nhypertension. We show how we make the fluids model tractable by reducing the\nparameter dimension from a 55D to a 5D problem. The Delayed Rejection Adaptive\nMetropolis (DRAM) algorithm, coupled with constraint nonlinear optimization is\nsuccessfully used to learn the parameter values and quantify the uncertainty in\nthe parameter estimates. To accommodate for different magnitudes of the\nparameter values, we introduce an improved parameter scaling technique in the\nDRAM algorithm. Formal convergence diagnostics are employed to check for\nconvergence of the Markov chains. Additionally, we perform model selection\nusing different information criteria, including Watanabe Akaike Information\nCriteria.\n