https://blog.csdn.net/mao_feng/article/details/78939864
In real life, we have encountered a small amount of sample labels, and a large number of samples without labels, how to do this deal with it?
Method 1: Migration finetune learning
Looking for similar universal data set to train the network, by modifying the rear layer 2 or layer 3 network, migration study done, to fine tune the parameters of the network, thus training the model.
Method 2: