Remote Sensing Image Classification with Few Labeled Data Using Semisupervised Learning

Table 2

Training parameters.

Parameter	Value
Number of labeled samples in a minibatch	8
Number of unlabeled samples in a minibatch	32
Initial learning rate	, being halved every iterations
Weight of unsupervised loss	1
Confidence threshold	0.5
Softmax thermal parameter	0.5
TSA strategy	Logarithmic form
Maximum number of iterations	25000