Surrogate-Assisted Optimization and Federated Learning Advances

The field of multi-objective optimization is seeing a shift towards the use of surrogate-assisted methods, which aim to improve the efficiency of optimization algorithms by augmenting the objective functions with data-driven models. These methods have been shown to achieve better solutions than conventional algorithms for a limited budget of function evaluations. Additionally, federated learning is becoming increasingly important, with a focus on balancing communication efficiency, model performance, and privacy protection. Researchers are exploring new frameworks and algorithms that can efficiently navigate the Pareto-optimal solution space and provide optimal balances between accuracy, communication efficiency, and privacy. Noteworthy papers include:

  • A Comparison-Relationship-Surrogate Evolutionary Algorithm for Multi-Objective Optimization, which introduces a new type of surrogate model that predicts domination relationships between candidate solutions.
  • A Generalized Meta Federated Learning Framework with Theoretical Convergence Guarantees, which presents a generalized framework for meta federated learning with theoretical convergence guarantees.

Sources

A Comparison-Relationship-Surrogate Evolutionary Algorithm for Multi-Objective Optimization

Convergence Analysis of Asynchronous Federated Learning with Gradient Compression for Non-Convex Optimization

Clustering-Based Evolutionary Federated Multiobjective Optimization and Learning

A Generalized Meta Federated Learning Framework with Theoretical Convergence Guarantees

The First Theoretical Approximation Guarantees for the Non-Dominated Sorting Genetic Algorithm III (NSGA-III)

Multi-start Optimization Method via Scalarization based on Target Point-based Tchebycheff Distance for Multi-objective Optimization

Built with on top of