Enhancing Semi-Supervised Learning via Representative and Diverse Sample Selection

However, we observe that how to select samples for labelling also significantly impacts performance, particularly under extremely low-budget settings. The sample selection task in SSL has been under-explored for a long time.