Model Compression - Cornell University

ModelCompressionCristianBucil hundredsor thousandsof base-level classi , thespacerequiredto storethismany clas-si ers,andthetimerequiredto executethemat run-time,prohibitstheirusein applicationswheretestsetsarelarge( ),wherestoragespaceis ata premium( ),andwherecomputationalpower is limited( ).We present a method for\compressing"large,complexensemblesin to smaller,fastermodels,usuallywith-outsign i cant lossin Subject [PatternRe-cognition]:Models{ :Algorithms,Experimentation,Measure-ment , Performance, :SupervisedLearning, a collectionof modelswhosepredictionsarecombinedby weightedaveragingor beenthefocusof signi cant research in thepastdecade,anda variety of ensemblemethods have knownensemblemethods includebagging[2],boosting[14],randomfor ests[3],Bayesianaveraging[9]andstacking[ 17].Much of theinterestin ensemblemethodshasbeenfueledby theirexcellent ,however,have onedisadvantagethatoftenis overlooked:many ensemblemethods unusableforapplicationswithlim-itedmemor y, storagespace,or computationalpower such asportabledevicesor sensornetworks,andforapplicationsinwhich ,forexam-ple,boosteddecisiontrees,bagged decisiontreesor thousandsof decisiontrees,each of which mustbe stored,andexecutedat run-timeto make singletreeisfast,butexecutinga thousandtreesis digitalorhardcopiesofallorpartofthiswork forpersonalorclassroomuseis copy otherwise,torepublish,topostonserversort oredistributetolists,requirespriorspecif icpermissi}

Given example e and its closest other example e0, the val-ues for each noncontinuous attribute are swapped between e and e0 with probability p and are left unchanged with probability 1 p. For each continuous attribute a, with. TRUE DIST RANDOM NBE MUNGE Figure 1: Synthetic data generated for a simple 2D

Loading..

Tags:

  Model, Compression, V alue, Model compression

Information

Related search queries