1 paper · 1 filter
Lukas Miklautz, Timo Klein, Kevin Sidak +5
This work investigates an important phenomenon in centroid-based deep clustering (DC) algorithms: Performance quickly saturates after a period of rapid early gains. Practitioners c…