Distributed Large-Scale Linear Programming
UC San Diego inventors have come up with a way to efficiently parallelize linear programming solvers using a fast-converging iterative method. The invention compares favorably against traditional conjugate-gradient methods (which are difficult to parallelize) and traditional iterative solvers (which are slow to converge to a solution). The method enables improved performance for the following industrial applications: Business Administration Product mix planningDistribution networksTruck routingStaff schedulingFinancial portfolios optimizationProfit maximizationCorporate restructuring Telecommunications Call routingNetwork designDecoding error-correcting codesInternet trafficNon-linear sampling (compressed sensing) Transportation Vehicle/packages routing (FedEx, UPS) Manufacturing VLSI chip board manufacturing Machine LearningSocial NetworksStatistical Software Packages
Tech ID/UC Case 19876/2009-243-0 Related Cases 2009-243-0
美国
