TY - JOUR

T1 - Dimensionality reduction of collective motion by principal manifolds

AU - Gajamannage, Kelum

AU - Butail, Sachit

AU - Porfiri, Maurizio

AU - Bollt, Erik M.

N1 - Funding Information:
Kelum Gajamannage and Erik M. Bollt have been supported by the National Science Foundation under grant no. CMMI-1129859 . Sachit Butail and Maurizio Porfiri have been supported by the National Science Foundation under grant no. CMMI-1129820 .

PY - 2015/1/15

Y1 - 2015/1/15

N2 - While the existence of low-dimensional embedding manifolds has been shown in patterns of collective motion, the current battery of nonlinear dimensionality reduction methods is not amenable to the analysis of such manifolds. This is mainly due to the necessary spectral decomposition step, which limits control over the mapping from the original high-dimensional space to the embedding space. Here, we propose an alternative approach that demands a two-dimensional embedding which topologically summarizes the high-dimensional data. In this sense, our approach is closely related to the construction of one-dimensional principal curves that minimize orthogonal error to data points subject to smoothness constraints. Specifically, we construct a two-dimensional principal manifold directly in the high-dimensional space using cubic smoothing splines, and define the embedding coordinates in terms of geodesic distances. Thus, the mapping from the high-dimensional data to the manifold is defined in terms of local coordinates. Through representative examples, we show that compared to existing nonlinear dimensionality reduction methods, the principal manifold retains the original structure even in noisy and sparse datasets. The principal manifold finding algorithm is applied to configurations obtained from a dynamical system of multiple agents simulating a complex maneuver called predator mobbing, and the resulting two-dimensional embedding is compared with that of a well-established nonlinear dimensionality reduction method.

AB - While the existence of low-dimensional embedding manifolds has been shown in patterns of collective motion, the current battery of nonlinear dimensionality reduction methods is not amenable to the analysis of such manifolds. This is mainly due to the necessary spectral decomposition step, which limits control over the mapping from the original high-dimensional space to the embedding space. Here, we propose an alternative approach that demands a two-dimensional embedding which topologically summarizes the high-dimensional data. In this sense, our approach is closely related to the construction of one-dimensional principal curves that minimize orthogonal error to data points subject to smoothness constraints. Specifically, we construct a two-dimensional principal manifold directly in the high-dimensional space using cubic smoothing splines, and define the embedding coordinates in terms of geodesic distances. Thus, the mapping from the high-dimensional data to the manifold is defined in terms of local coordinates. Through representative examples, we show that compared to existing nonlinear dimensionality reduction methods, the principal manifold retains the original structure even in noisy and sparse datasets. The principal manifold finding algorithm is applied to configurations obtained from a dynamical system of multiple agents simulating a complex maneuver called predator mobbing, and the resulting two-dimensional embedding is compared with that of a well-established nonlinear dimensionality reduction method.

KW - Algorithm

KW - Collective behavior

KW - Dimensionality reduction

KW - Dynamical systems

UR - http://www.scopus.com/inward/record.url?scp=84910089515&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84910089515&partnerID=8YFLogxK

U2 - 10.1016/j.physd.2014.09.009

DO - 10.1016/j.physd.2014.09.009

M3 - Article

AN - SCOPUS:84910089515

VL - 291

SP - 62

EP - 73

JO - Physica D: Nonlinear Phenomena

JF - Physica D: Nonlinear Phenomena

SN - 0167-2789

ER -