Approximation Analysis of Gradient Descent Algorithm for Bipartite Ranking

Table 1

The values of parameters for different convex losses.

Loss function	𝐶0	𝑞	Δ∗	𝛼
𝜙(𝑡)=𝑡^2	1	2	1+12𝜅^2	𝜃min{(1/2)(𝜃−3𝛾),𝛾𝛽}
𝜙(𝑡)=log(1+𝑒^−𝑡)	2	1	1+11𝜅^2	(𝛾+𝜃)min{𝜃−𝛾,𝛾−𝛽}
𝜙(𝑡)=(1−𝑡)+	1	1	1+8𝜅^2	𝜃min{(1/2)(𝜃−3𝛾),𝛾𝛽}