Table 3.

Explicit transfer learning methods.

	Approach	References	Paradigm	Summary
Non-parametric alignment	MMD	Hang et al., 2019	Motor imagery	Minimized MMD in a feature level and introduced CDFL
		Chai et al., 2016	Emotion	Minimized MMD in a feature level and trained AE and classifier separately
	KLD	Zhang et al., 2017	Sleep	Minimized KLD in a feature level and trained with classifier in an end-to-end manner
	EA	Kostas and Rudzicz, 2020	Multi	Constrained that the mean covariance matrix becomes an identity matrix in a raw data level
Adversarial learning	A-cVAE (Wang Y. et al., 2018)	Özdenizci et al., 2019	Motor imagery	Added an adversarial network to cVAE, and trained cVAE and classifier separately
	DANN (Ganin et al., 2016)	Özdenizci et al., 2020	Motor imagery	Devised DANN by exploiting various CNN-based architectures as their feature extractor
		Zhao H. et al., 2020		Added center loss for target to minimize intra-class compactness and maximize inter-class separability
		Tang and Zhang, 2020		Fed output of a classifier into a domain discriminator
		Jeon et al., 2019		Selected source based on resting-state EEG signals
		Wei et al., 2020b	RSVP	Selected sources based on a ranking of performances in subject-specific classifiers
		Wang et al., 2021	Emotion	Selected sources based on a ranking of performances in subject-specific classifiers and devised centroid alignment loss
		Nasiri and Clifford, 2020	Sleep	Estimated attention maps using channel-wise domain discriminators
		Ma et al., 2019	Drowsy	Trained additional parameters capturing subject-specific features