Parameter Space Compression Underlies Emergent Theories and Predictive Models

• Published in: Science (American Association for the Advancement of Science) 2013-11, Vol.342 (6158), p.604-607
• Main Authors: Machta, Benjamin B., Chachra, Ricky, Transtrum, Mark K., Sethna, James P.
• Format: Article
• Language: English
• Subjects: Computational methods in statistical physics and nonlinear dynamics; Critical points; Eigenvalues; Eigenvectors; Exact sciences and technology; Information Theory; Ising model; Modeling; Parameter estimation; Parametric models; Particle density; Physics; Predictive modeling; Science; Statistical models; Statistical physics, thermodynamics, and nonlinear dynamical systems
• ISSN: 0036-8075; 1095-9203
• DOI: 10.1126/science.1238723

The microscopically complicated real world exhibits behavior that often yields to simple yet quantitatively accurate descriptions. Predictions are possible despite large uncertainties in microscopic parameters, both in physics and in multiparameter models in other areas of science. We connect the two by analyzing parameter sensitivities in a prototypical continuum theory (diffusion) and at a self-similar critical point (the Ising model). We trace the emergence of an effective theory for long-scale observables to a compression of the parameter space quantified by the eigenvalues of the Fisher Information Matrix. A similar compression appears ubiquitously in models taken from diverse areas of science, suggesting that the parameter space structure underlying effective continuum and universal theories in physics also permits predictive modeling more generally.