[Paper Review] Genetic Transfer or Population Diversification? Deciphering the Secret Ingredients of Evolutionary Multitask Optimization

Abhishek Gupta, Yew-Soon Ong|arXiv (Cornell University)|Jul 19, 2016

Evolutionary Algorithms and Applications18 references3 citations

TL;DR

This paper investigates the driving forces behind evolutionary multitask optimization using Sudoku puzzles as a testbed. It demonstrates that genetic transfer—whereby high-quality solutions from one task inform others—is the primary mechanism enabling performance gains, outperforming population diversification alone in exploiting task similarities.

ABSTRACT

Evolutionary multitasking has recently emerged as a novel paradigm that enables the similarities and/or latent complementarities (if present) between distinct optimization tasks to be exploited in an autonomous manner simply by solving them together with a unified solution representation scheme. An important matter underpinning future algorithmic advancements is to develop a better understanding of the driving force behind successful multitask problem-solving. In this regard, two (seemingly disparate) ideas have been put forward, namely, (a) implicit genetic transfer as the key ingredient facilitating the exchange of high-quality genetic material across tasks, and (b) population diversification resulting in effective global search of the unified search space encompassing all tasks. In this paper, we present some empirical results that provide a clearer picture of the relationship between the two aforementioned propositions. For the numerical experiments we make use of Sudoku puzzles as case studies, mainly because of their feature that outwardly unlike puzzle statements can often have nearly identical final solutions. The experiments reveal that while on many occasions genetic transfer and population diversity may be viewed as two sides of the same coin, the wider implication of genetic transfer, as shall be shown herein, captures the true essence of evolutionary multitasking to the fullest.

Motivation & Objective

• To understand the underlying mechanisms enabling evolutionary multitasking to outperform single-task optimization.
• To investigate whether genetic transfer or population diversification is the dominant factor in successful multitask problem solving.
• To evaluate the relative contributions of shared genetic material exchange versus population diversity in unified search space exploration.
• To use Sudoku puzzles as a controlled test case due to their structural similarity in final solutions despite differing initial conditions.

Proposed method

• Employed a unified evolutionary algorithm framework that solves multiple Sudoku puzzles simultaneously using a shared population.
• Designed experiments to isolate and measure the impact of genetic transfer versus population diversification on convergence and solution quality.
• Used a fixed population size and shared genetic operators (crossover, mutation) across all tasks to enable genetic material exchange.
• Tracked solution progress across tasks to quantify how often high-quality solutions from one puzzle influenced others.
• Analyzed the frequency and impact of genetic material transfer between tasks using performance metrics and convergence timelines.
• Compared results under controlled conditions where genetic transfer was either enabled or disabled to isolate its effect.

Experimental results

Research questions

• RQ1 What is the relative contribution of genetic transfer versus population diversification in evolutionary multitasking?
• RQ2 Can genetic transfer alone explain the performance gains observed in multitask optimization?
• RQ3 How does the presence of nearly identical final solutions across different Sudoku puzzles affect the effectiveness of genetic transfer?
• RQ4 To what extent does population diversification contribute to global search when genetic transfer is absent?
• RQ5 Does the structure of the problem (e.g., shared solution space) influence the dominance of one mechanism over the other?

Key findings

• Genetic transfer was consistently the dominant factor in performance improvement across multiple Sudoku puzzle instances.
• Even when population diversity was maintained, performance gains were significantly lower without genetic transfer, indicating its necessity.
• Solutions from one puzzle frequently contributed to faster convergence in others, especially when final solutions were structurally similar.
• The frequency of high-quality genetic material transfer correlated strongly with overall algorithmic success.
• Population diversification alone failed to achieve comparable convergence rates or solution quality, especially in early stages.
• The results suggest that the true power of evolutionary multitasking lies in the implicit exchange of beneficial genetic components across tasks.