On Martingales, Causality, Identifiability and Model Selection
Publikation: Bog/antologi/afhandling/rapport › Ph.d.-afhandling › Forskning
Standard
On Martingales, Causality, Identifiability and Model Selection. / Sokol, Alexander.
Department of Mathematical Sciences, Faculty of Science, University of Copenhagen, 2013. 246 s.Publikation: Bog/antologi/afhandling/rapport › Ph.d.-afhandling › Forskning
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - BOOK
T1 - On Martingales, Causality, Identifiability and Model Selection
AU - Sokol, Alexander
PY - 2013
Y1 - 2013
N2 - Ornstein-Uhlenbeck SDEs, where explicit calculations may be made for the postinterventiondistributions.Chapter 9 concerns identiability of the mixing matrix in ICA. It is a well-knownresult that identiability of the mixing matrix depends crucially on whether theerror distributions are Gaussian or not. We attempt to elucidate what happens inthe case where the error distributions are close to but not exactly Gaussian.Finally, Chapter 10 discusses degrees of freedom in nonlinear regression. Our motivatingproblem is that of L1-constrained and L1-penalized estimation in nonlinearregression. Our objective is to obtain results leading to the calculation of the degreesof freedom of an estimator in order to enable sparse model selection by optimalchoice of the penalization parameter. We prove two results related to the degrees offreedom, one theoretical result for constrained estimation, and one more practicallyapplicable for L1-penalized estimation.
AB - Ornstein-Uhlenbeck SDEs, where explicit calculations may be made for the postinterventiondistributions.Chapter 9 concerns identiability of the mixing matrix in ICA. It is a well-knownresult that identiability of the mixing matrix depends crucially on whether theerror distributions are Gaussian or not. We attempt to elucidate what happens inthe case where the error distributions are close to but not exactly Gaussian.Finally, Chapter 10 discusses degrees of freedom in nonlinear regression. Our motivatingproblem is that of L1-constrained and L1-penalized estimation in nonlinearregression. Our objective is to obtain results leading to the calculation of the degreesof freedom of an estimator in order to enable sparse model selection by optimalchoice of the penalization parameter. We prove two results related to the degrees offreedom, one theoretical result for constrained estimation, and one more practicallyapplicable for L1-penalized estimation.
UR - https://soeg.kb.dk/permalink/45KBDK_KGL/fbp0ps/alma99122333045405763
M3 - Ph.D. thesis
SN - 978-87-7078-386-6
BT - On Martingales, Causality, Identifiability and Model Selection
PB - Department of Mathematical Sciences, Faculty of Science, University of Copenhagen
ER -
ID: 108562239