P-GAMs have similar scaling with the number of input dimensions as traditional GLMs

Feb-9-2026, 09:44:22 GMT–Neural Information Processing Systems 

We need some form of group sparsity and traditional ARD can't do that.