Statement
Here E=R1 with its Euclidean norm, and the fixed penalty has domain D=E and total value ρ(z)=(μ/2)∥z∥2. For a real L, a pair of total functions f:E→R and g:E→E is L-smooth convex here when f is convex on all of E, f has gradient g(z) at every z∈E, and ∥g(a)−g(b)∥≤L∥a−b∥ for every a,b∈E. For real γ, natural n, and a sequence x:N→E, the trajectory condition through n means that xi∈D for every natural i≤n, and that, for every natural i<n and every y∈D, ρ(xi+1)+∥xi+1−(xi−γg(xi))∥2/(2γ)≤ρ(y)+∥y−(xi−γg(xi))∥2/(2γ). No condition is imposed on indices above n. Set Δn=f(x0)+ρ(x0)−f(xn)−ρ(xn) and Rn(γ)=∥xn−1−xn∥2/γ2, where n−1 is truncated to 0 at n=0 and real division by zero is 0. For N∈N (including 0) and h,q∈R, put b=1+hq, define AN(h,q)=2/[h(2N−h)] if q=0 and AN(h,q)=2q/[b2N−1−h2q] otherwise, define BN(h,q)=2(1+q)(1−h)2N−2/[b2N−(1−h)2N], and define CN(h,q)=2/[h(2−h+hq)] when N=1 and CN(h,q)=max{AN(h,q),BN(h,q)} otherwise. The exponent 2N−2 in BN uses subtraction in N, so it is 0 when N=0 or N=1. Every division is division in R, with division by zero equal to zero. For every real L,μ,h and natural N with L>0, μ≥0, 0<h<2, and N≥1, take E=R1, D=E, and ρ(z)=(μ/2)∥z∥2. There exist total f:E→R, g:E→E, and x:N→E for which f,g are L-smooth convex, the trajectory condition holds through N with γ=h/L, ΔN>0, and CN(h,μ/L)=RN(h/L)/(LΔN); no quadratic form for f is required.