A survey on heterogeneous transfer learning

Journal of Big Data

Table 1 Surveyed HTL methods which require limited target labels

Methods	Characteristics	Sections
COTL [13]	Online tasks, multiview ensemble	“COTL, OHTWC”
OHTWC [19]	Requires co-occurrence data, online tasks, weighted ensemble	“COTL, OHTWC”
HFA [23]	Symmetric transformation w/augmentation	“HFA”
HMCA [29]	Requires source classifier, model-transfer	“HMCA”
SHFR [31]	Asymmetric, multi-class	“SHFR”
TTI [36]	Requires co-occurrence data, translator, image classification	“TTI”
TLRisk [37]	Requires co-occurrence data, translator, Markov chain, risk minimization	“TLRisk”
ARC-t [28]	Asymmetric, image classification, kernel methods	“ARC-t”
SHDA-RF [41]	Asymmetric, random forest and label pivots	“SHDA-RF”
IFSR [12]	Asymmetric, domain-independent metafeatures to compute similarity	“FSR (IFSR, UFSR, ELFSR)”
ELFSR [12]	Asymmetric, FSR ensemble, multi-source, stacking method	“FSR (IFSR, UFSR, ELFSR)”
SMVCCAE [64]	Symmetric, multi-view ensemble, CCA analysis	“SMVCCAE, SSMVCCAE”