1: \begin{abstract}

2: We develop a novel primal-dual algorithm to solve a class of nonsmooth and nonlinear compositional convex minimization problems, which covers many existing and brand-new models as special cases.

3: Our approach relies on a combination of a new nonconvex potential function, Nesterov's accelerated scheme, and an adaptive parameter updating strategy.

4: Our algorithm is single-loop and has low per-iteration complexity.

5: Under only general convexity and mild assumptions, our algorithm achieves $\BigO{1/k}$ convergence rates through three different criteria: primal objective residual, dual objective residual, and primal-dual gap, where $k$ is the iteration counter.

6: Our rates are both \textit{ergodic} (i.e., on an averaging sequence) and \textit{non-ergodic} (i.e., on the last-iterate sequence).

7: These convergence rates can be accelerated up to $\BigO{1/k^2}$ if only one objective term is strongly convex (or equivalently, its conjugate is $L$-smooth).

8: To the best of our knowledge, this is the first algorithm achieving optimal rates on the primal last-iterate sequence for nonlinear compositional convex minimization.

9: As a by-product, we specify our algorithm to solve a general convex cone constrained program with both ergodic and non-ergodic rate guarantees.

10: %Our new algorithm  achieves both ergodic and non-ergodic optimal convergence rates on both the primal objective residual and the primal feasibility violation for four cases.

11: We test our algorithms and compare them with two recent methods on a binary classification and a convex-concave game model.

12: 

13: %%%% Keywords.

14: \vspace{1ex}

15: \noindent\textbf{Keywords:}

16: Nonlinear compositional convex minimization; 

17: convex-concave minimax problem;

18: primal-dual algorithm; 

19: optimal convergence rate; 

20: Nesterov's accelerated scheme; 

21: convex cone constrained program.

22: \end{abstract}

23: