Instance Space Analysis Toolbox v0.9.1

Instance Space Analysis Toolbox

Objective algorithm testing through the instance space of a problem

The Instance Space Analysis (ISA) Toolbox relates the structural features of problem instances to the performance of a portfolio of algorithms. It projects every instance into a 2D or 3D instance space, shows where each algorithm performs well (its footprint), predicts which algorithm to use for a new instance, and estimates where in the space no instances have been observed yet.

The toolbox implements the methodology in Smith-Miles & Muñoz (2023), including the 3D extension (ISA3) of Simpson et al. (2025). It runs in MATLAB R2025a or later and powers the MATILDA web platform.

The ISA pipeline: INIT and PRELIM prepare the metadata, SIFTED selects features, PILOT projects them, then CLOISTER, PYTHIA and TRACE work in the projected space. INITread metadata PRELIMlabel & normalise SIFTEDselect features PILOTproject to 2D/3D CLOISTERempirical bounds PYTHIApredict algorithms TRACEfootprints OutputsCSV · PNG · .mat

The ISA pipeline. InstanceSpace runs the stages in this order; each stage is also a standalone function.

Get Started

Getting Started

Build and explore an instance space from the bundled reference data in a few lines.

Stage-by-Stage Walkthrough

Run each stage separately and inspect what it produces.

Metadata File Format

Prepare metadata.csv for your own problem domain.

Options Reference

Every field of opts and options.json, with defaults.

Functions and Classes

InstanceSpace

Build, explore, save, load and plot an instance space. The main entry point.

Pipeline Stages

INIT, PRELIM, FILTER, SIFTED, PILOT, PILOTviewpoint, CLOISTER, PYTHIA, TRACE.

Utilities

Option defaults and validation, classifier registry, legacy-model migration.

Output

Write CSV files and figures, and restore 3D viewpoints.

Topics

Installation

Clone or download the repository, then add it to the MATLAB path from its root folder:

cd InstanceSpace
addpath(pwd)     % the class, the wrappers, and info.xml for the Help browser
startup          % adds core/, utils/, output/ and deprecated/

Required products: MATLAB R2025a or later with the Statistics and Machine Learning, Optimization, Global Optimization, Parallel Computing, and Financial toolboxes.

With the repository root on the path, this documentation also appears in the MATLAB Help browser under Supplemental Software.

References

  • Smith-Miles, K. & Muñoz, M.A. (2023). Instance Space Analysis for Algorithm Testing. ACM Computing Surveys, 55(12), Article 255. https://doi.org/10.1145/3572895
  • Simpson, C., Muñoz, M.A., Kandanaarachchi, S. & Campello, R.J.G.B. (2025). ISA3: A 3-dimensional expansion of Instance Space Analysis. Machine Learning, 114, 240. https://doi.org/10.1007/s10994-025-06871-5

Report a documentation issue · Source on GitHub

Copyright © 2026 Mario Andrés Muñoz Acosta and contributors, School of Computing and Information Systems, The University of Melbourne. Released under the PolyForm Noncommercial License 1.0.0.