Hands-On Meta Learning with Python
Metalearningisanexcitingresearchtrendinmachinelearning,whichenablesamodeltounderstandthelearningprocess.UnlikeotherMLparadigms,withmetalearningyoucanlearnfromsmalldatasetsfaster.Hands-OnMetaLearningwithPythonstartsbyexplainingthefundamentalsofmetalearningandhelpsyouunderstandtheconceptoflearningtolearn.Youwilldelveintovariousone-shotlearningalgorithms,likesiamese,prototypical,relationandmemory-augmentednetworksbyimplementingtheminTensorFlowandKeras.Asyoumakeyourwaythroughthebook,youwilldiveintostate-of-the-artmetalearningalgorithmssuchasMAML,Reptile,andCAML.YouwillthenexplorehowtolearnquicklywithMeta-SGDanddiscoverhowyoucanperformunsupervisedlearningusingmetalearningwithCACTUs.Intheconcludingchapters,youwillworkthroughrecenttrendsinmetalearningsuchasadversarialmetalearning,taskagnosticmetalearning,andmetaimitationlearning.Bytheendofthisbook,youwillbefamiliarwithstate-of-the-artmetalearningalgorithmsandabletoenablehuman-likecognitionforyourmachinelearningmodels.
·4萬字