Skip to main content

Table 2 Homogeneous transfer learning approaches surveyed in “Homogeneous transfer learning” section listing different characteristics of each approach

From: A survey of transfer learning

Approach Transfer category Source data Target data Multiple sources Generic solution Negative transfer
CP-MDA [14] Parameter Labeled Limited labels  
2SW-MDA [14] Instance Labeled Unlabeled  
FAM [22] Asymmetric feature Labeled Limited labels  
DTMKL [27] Asymmetric feature Labeled Unlabeled    
JDA [69] Asymmetric feature Labeled Unlabeled    
ARTL [68] Asymmetric feature Labeled Unlabeled    
TCA [87] Symmetric feature Labeled Unlabeled    
SFA [83] Symmetric feature Labeled Limited labels  
SDA [41] Symmetric feature Labeled Unlabeled    
GFK [42] Symmetric feature Labeled Unlabeled  
DCP [106] Symmetric feature Labeled Unlabeled    
TCNN [81] Symmetric feature Labeled Limited labels    
MMKT [114] Parameter Labeled Limited labels
DSM [28] Parameter Labeled Unlabeled  
MsTrAdaBoost [138] Instance Labeled Limited labels
TaskTrAdaBoost [138] Parameter Labeled Limited labels
RAP [62] Relational Labeled Unlabeled    
SSFE [132] Hybrid (instance and feature) Labeled Limited labels