TY - GEN
T1 - Geodesic least squares regression on the gaussian manifold with an application in astrophysics
AU - Verdoolaege, Geert
N1 - Publisher Copyright:
© 2017, Springer International Publishing AG.
PY - 2017
Y1 - 2017
N2 - We present a new regression method called geodesic least squares (GLS), which is particularly robust against data and model uncertainty. It is based on minimization of the Rao geodesic distance on a probabilistic manifold. We apply GLS to Tully-Fisher scaling of the total baryonic mass vs. the rotation velocity in disk galaxies and we show the excellent robustness properties of GLS for estimating the coefficients and the tightness of the scaling.
AB - We present a new regression method called geodesic least squares (GLS), which is particularly robust against data and model uncertainty. It is based on minimization of the Rao geodesic distance on a probabilistic manifold. We apply GLS to Tully-Fisher scaling of the total baryonic mass vs. the rotation velocity in disk galaxies and we show the excellent robustness properties of GLS for estimating the coefficients and the tightness of the scaling.
KW - Geodesic least squares
KW - Rao geodesic distance
KW - Robust regression
KW - Tully-Fisher scaling
UR - http://www.scopus.com/inward/record.url?scp=85033718879&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-68445-1_72
DO - 10.1007/978-3-319-68445-1_72
M3 - Conference contribution
AN - SCOPUS:85033718879
SN - 9783319684444
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 621
EP - 628
BT - Geometric Science of Information - 3rd International Conference, GSI 2017, Proceedings
A2 - Nielsen, Frank
A2 - Barbaresco, Frederic
A2 - Nielsen, Frank
PB - Springer
T2 - 3rd International Conference on Geometric Science of Information, GSI 2017
Y2 - 7 November 2017 through 9 November 2017
ER -