Abstract:This paper considers distributed optimization problems, where each agent cooperatively minimizes the sum of local objective functions through the communication with its neighbors. The widely adopted distributed gradient method in solving this problem suffers from slow convergence rates, which motivates us to incorporate the second-order information of the objective functions. However, the challenge arises from the unique structure of the inverse of the Hessian matrix, which prevents the direct distributed implementation of the second-order method. We overcome this challenge by proposing a novel optimization framework. The key idea is to transform the distributed optimization problem into an optimal control problem. Using Pontryagin's maximum principle and the associated forward-backward difference equations (FBDEs), we derive a new distributed optimization algorithm that incorporates the second-order information without requiring the computation of the inverse of the Hessian matrix. Furthermore, the superlinear convergence of the proposed algorithm is proved under some mild assumptions. Finally, we also propose a variant of the algorithm to balance the number of iterations and communication.

What problem does this paper attempt to address?

This paper aims to solve the problem of slow convergence rate in distributed optimization problems. Specifically, each agent minimizes the sum of local objective functions through communication and cooperation with neighbors. The existing distributed gradient methods have the problem of slow convergence speed when solving this problem, which prompts researchers to consider introducing second - order information of the objective function to accelerate convergence. However, due to the unique structure of the inverse of the Hessian matrix, directly implementing second - order methods in a distributed environment is challenging. To this end, the author proposes a new optimization framework. By transforming the distributed optimization problem into an optimal control problem and using the Pontryagin maximum principle and related forward - backward difference equations (FBDEs), a new distributed optimization algorithm is derived. This algorithm can incorporate second - order information without calculating the inverse of the Hessian matrix, thereby achieving super - linear convergence. ### Main contributions: 1. **New optimization framework**: An innovative optimization framework is proposed by transforming the distributed optimization problem into an optimal control problem. 2. **Avoiding the calculation of the inverse of the Hessian matrix**: The new algorithm avoids the calculation of the inverse of the Hessian matrix through a special optimization structure, significantly reducing the computational complexity. 3. **Proof of super - linear convergence**: Under certain assumptions, the super - linear convergence of the proposed algorithm is proved. 4. **Balancing the number of iterations and communication**: An algorithm variant is proposed to balance the number of iterations and the amount of communication. ### Key technical points: - **Pontryagin maximum principle**: Used to derive the new optimization algorithm. - **Forward - backward difference equations (FBDEs)**: Used to describe the dynamics of the algorithm. - **Fusion of second - order information**: Incorporate second - order information without calculating the inverse of the Hessian matrix through a special optimization structure. ### Application background: This research has broad application prospects in fields such as sensor networks, distributed control, and machine learning, especially in scenarios where multiple agents need to cooperate for optimization. By increasing the convergence speed of the distributed optimization algorithm, the efficiency and performance of these applications can be significantly improved.

Distributed Optimization Algorithm with Superlinear Convergence Rate

An Initialization-Free Continuous-Time Distributed Algorithm for Convex Optimization Problem

Distributed Algorithms for Composite Optimization: Unified Framework and Convergence Analysis

Linearly Convergent Second-Order Distributed Optimization Algorithms

Distributed adaptive Newton methods with global superlinear convergence

Achieving Globally Superlinear Convergence for Distributed Optimization with Adaptive Newton Method

Distributed Stochastic Algorithm for Global Optimization in Networked System

Distributed Adaptive Newton Methods with Globally Superlinear Convergence

Accelerated Primal-Dual Algorithms for Distributed Smooth Convex Optimization over Networks

Distributed Zeroth-Order Optimization: Convergence Rates That Match Centralized Counterpart

Convergence in High Probability of Distributed Stochastic Gradient Descent Algorithms

A Unified Algorithmic Framework for Distributed Composite Optimization.

Distributed Nonconvex Optimization: Gradient-free Iterations and $ε$-Globally Optimal Solution

Distributed proximal‐gradient algorithms for nonsmooth convex optimization of second‐order multiagent systems

Distributed Optimization Based on Gradient-tracking Revisited: Enhancing Convergence Rate via Surrogation

A Unified Contraction Analysis of a Class of Distributed Algorithms for Composite Optimization

A Push-Pull Gradient Method for Distributed Optimization in Networks.

Distributed stochastic projection-free solver for constrained optimization

Exponential convergence of distributed optimization for heterogeneous linear multi-agent systems

A Communication-Efficient Stochastic Gradient Descent Algorithm for Distributed Nonconvex Optimization

Linear Convergence of Distributed Aggregative Optimization with Coupled Inequality Constraints