Island-Based Evolutionary Computation with Diverse Surrogates and Adaptive Knowledge Transfer for High-Dimensional Data-Driven Optimization

TL;DR

The paper tackles the challenge of expensive evaluations in high-dimensional, large-scale optimization by introducing DSKT-DDEA, an offline, island-based evolutionary algorithm that uses diverse surrogate models and adaptive knowledge transfer. It combines intra-island semi-supervised surrogate fine-tuning with inter-island migration driven by learned attractiveness and model-discrepancy factors to maintain diversity and avoid premature convergence. Through extensive experiments on 1000-dimensional BBOB and CEC2010 benchmarks, the method demonstrates competitive performance and scalable parallelism, supported by an open-source implementation. The results show that combining diverse surrogates with data-informed migration markedly improves optimization under data scarcity, offering a practical approach for offline, large-scale optimization problems.

Abstract

In recent years, there has been a growing interest in data-driven evolutionary algorithms (DDEAs) employing surrogate models to approximate the objective functions with limited data. However, current DDEAs are primarily designed for lower-dimensional problems and their performance drops significantly when applied to large-scale optimization problems (LSOPs). To address the challenge, this paper proposes an offline DDEA named DSKT-DDEA. DSKT-DDEA leverages multiple islands that utilize different data to establish diverse surrogate models, fostering diverse subpopulations and mitigating the risk of premature convergence. In the intra-island optimization phase, a semi-supervised learning method is devised to fine-tune the surrogates. It not only facilitates data argumentation, but also incorporates the distribution information gathered during the search process to align the surrogates with the evolving local landscapes. Then, in the inter-island knowledge transfer phase, the algorithm incorporates an adaptive strategy that periodically transfers individual information and evaluates the transfer effectiveness in the new environment, facilitating global optimization efficacy. Experimental results demonstrate that our algorithm is competitive with state-of-the-art DDEAs on problems with up to 1000 dimensions, while also exhibiting decent parallelism and scalability. Our DSKT-DDEA is open-source and accessible at: https://github.com/LabGong/DSKT-DDEA.

Figures (9)

• Figure 1: RBFN structure.
• Figure 2: The overall framework of DSKT-DDEA.
• Figure 3: Example of source island and target island. Source island "$i$" can only migrate to the neighboring island in Von Neumann topology, and similarly, target island "$o$" can only receive immigrants from the neighboring islands in Von Neumann. The sum of migration probability of source island emitting is 1.
• Figure 4: Analyzing the accuracy of our diverse surrogates.
• Figure 5: Fine-tuning vs. no fine-tuning, comparison of RMSE of all surrogate models.