Search

Recent workat the U.S. Army CCDC Ground Vehicle Systems Center has shown that machine learning classifiers can quickly construct high-fidelity mobility maps. Training these classifiers, on the other hand, is still a challenge, since each data instance is labeled by performing a computationally intensive, physics-based simulation. In this paper we introduce an active learning framework, based on the query-by-bagging algorithm, that substantially reduces the number of simulations needed to train a classifier. Experimental results suggest that our sampling algorithm can train a neural network, with higher accuracy, using less than half the number of simulations when compared to random sampling. [ABSTRACT FROM AUTHOR]

A multiscale continuum model is constructed for a mechanosensitive (MS) channel gated by tension in a lipid bilayer membrane under stresses due to fluid flows. We illustrate that for typical physiological conditions vesicle hydrodynamics driven by a fluid flow may render the membrane tension sufficiently large to gate a MS channel open. In particular, we focus on the dynamic opening/ closing of a MS channel in a vesicle membrane under a planar shear flow and a pressure-driven flow across a constriction channel. Our modeling and numerical simulation results quantify the critical flow strength or flow channel geometry for intracellular transport through a MS channel. In particular, we determine the percentage of MS channels that are open or closed as a function of the relevant measure of flow strength. The modeling and simulation results imply that for fluid flows that are physiologically relevant and realizable in microfluidic configurations stress-induced intracellular transport across the lipid membrane can be achieved by the gating of reconstituted MS channels, which can be useful for designing drug delivery in medical therapy and understanding complicated mechanotransduction. [ABSTRACT FROM AUTHOR]