1 paper · 1 filter
Albert Saiapin, Kim Batselier
Many approximations were suggested to circumvent the cubic complexity of kernel-based algorithms, allowing their application to large-scale datasets. One strategy is to consider th…