Multitasking Evolutionary Algorithm Based on Adaptive Seed Transfer for Combinatorial Problem

Haoyuan Lv,Ruochen Liu

2023-08-24

Abstract:Evolutionary computing (EC) is widely used in dealing with combinatorial optimization problems (COP). Traditional EC methods can only solve a single task in a single run, while real-life scenarios often need to solve multiple COPs simultaneously. In recent years, evolutionary multitasking optimization (EMTO) has become an emerging topic in the EC community. And many methods have been designed to deal with multiple COPs concurrently through exchanging knowledge. However, many-task optimization, cross-domain knowledge transfer, and negative transfer are still significant challenges in this field. A new evolutionary multitasking algorithm based on adaptive seed transfer (MTEA-AST) is developed for multitasking COPs in this work. First, a dimension unification strategy is proposed to unify the dimensions of different tasks. And then, an adaptive task selection strategy is designed to capture the similarity between the target task and other online optimization tasks. The calculated similarity is exploited to select suitable source tasks for the target one and determine the transfer strength. Next, a task transfer strategy is established to select seeds from source tasks and correct unsuitable knowledge in seeds to suppress negative transfer. Finally, the experimental results indicate that MTEA-AST can adaptively transfer knowledge in both same-domain and cross-domain many-task environments. And the proposed method shows competitive performance compared to other state-of-the-art EMTOs in experiments consisting of four COPs.

Neural and Evolutionary Computing

What problem does this paper attempt to address?

The paper aims to address the application of Evolutionary Multi-task Optimization (EMTO) in Combinatorial Optimization Problems (COP). Specifically, the researchers propose a new Evolutionary Multi-task Algorithm based on Adaptive Seed Transfer (MTEA-AST) to tackle combinatorial optimization problems in a multi-task environment. The paper mainly focuses on the following issues:

1. **Negative Transfer in Multi-task Optimization**: In a multi-task environment, the target task is influenced by both positive and negative tasks, which can weaken the effect of positive transfer and lead to negative transfer. Therefore, dynamically capturing the similarity between tasks and adjusting the transfer intensity becomes key to solving multi-task optimization.

2. **Cross-domain Knowledge Transfer between Heterogeneous Problems**: When extending EMTO to combinatorial optimization problems, there are differences in dimensions, individual representation, and fitness calculation between different tasks. Direct cross-domain knowledge transfer can lead to failure or severe negative transfer.

The proposed MTEA-AST method addresses the above issues through the following strategies:

- **Dimension Unification Strategy**: Map individuals of different tasks to a unified dimension, introducing a greedy heuristic method to replace meaningless information in random encoding, thereby reducing negative transfer.

- **Adaptive Task Selection Strategy**: Design an adaptive task selection strategy based on similarity, dynamically calculating the similarity between different tasks and selecting appropriate source tasks for knowledge transfer based on these similarities.

- **Task Transfer Strategy**: Select excellent seeds from the source task and transfer them to the population responsible for the target task to suppress negative transfer caused by heterogeneous problems.

Experimental results show that MTEA-AST performs comparably to existing EMTO methods in both intra-domain and cross-domain multi-task environments, demonstrating its effectiveness and competitiveness.