Representation shapes optimizer success in multitasking piano fingering tasks

Diagnosing Optimizer-Landscape Interaction via Representation in Subset Selection Multitasking: Evidence from Piano Fingering

Neural and Evolutionary ComputingMachine Learning

Summary

Finding the best way to choose piano fingerings is tricky because computers use methods that explore many possibilities at once. This paper looks at how different ways of representing these choices in the computer affect how easy it is to find good fingerings across multiple related tasks. The authors show that how the problem is represented changes the 'landscape' the computer explores, which greatly impacts how well optimization methods work together and share knowledge. In short, the design of the problem representation matters more than just the optimization algorithm when solving several tasks together.

What this means in practice

  • For music software developers: Optimize piano fingering estimation tools by selecting representations that improve search and transfer between multiple fingerings tasks.
  • For multitask optimization engineers: Design multitasking algorithms with representation-focused approaches to enhance knowledge transfer and solution quality across different selection problems.

Tested on one dataset.

Authors

Ananda Phan Iman, Tingyang Wei, Chang Wook Ahn

Abstract

Continuous relaxations are commonly used as an encoding scheme for evolutionary multitasking optimization (EMTO), allowing heterogeneous tasks to be optimized within a unified search space. However, these continuous representations fundamentally reshape the induced fitness landscape, and their impact on optimization behavior and knowledge transfer remains underexplored. In this work, we bridge the gap by analyzing the interaction between the representation-induced fitness landscape and optimizer behavior through fitness landscape analysis. Using piano fingering estimation as a representative case of the subset selection problem, we introduce several encoding variants and investigate how their induced landscapes influence search dynamics and solution quality under EMTO. Our analysis indicates that transfer efficacy is strongly associated with changes in representation-induced landscape characteristics, rather than algorithmic factors alone, highlighting the representation design as the key factor of effective knowledge transfer in multitasking optimization.
View PDFOpen arXiv