Trendafilov, Nickolay T.
|DOI (Digital Object Identifier) Link:||http://doi.org/10.1016/S0167-739X(03)00043-8|
|Google Scholar:||Look up in Google Scholar|
In this paper, the well-known Procrustes problem is reconsidered. The usual least squares objective function is replaced by more robust one, based on a smooth approximation of the ℓ 1 matrix norm. This smooth approximation to the ℓ1 Procrustes problem is solved making use of the projected gradient method. The Procrustes problem with partially specified target is treated and solved as well. Several classical numerical examples from factor analysis (well-known with their least squares Procrustes solutions) are solved with respect to the smooth approximation of the ℓ 1 matrix norm goodness-of-fit measure.
|Item Type:||Journal Article|
|Copyright Holders:||2003 Elsevier Science B.V.|
|Keywords:||fitting configurations; constrained optimization; dynamical system on manifolds; descent flows; optimality conditions; factor-pattern and reference-structure; partially specified target|
|Academic Unit/Department:||Faculty of Science, Technology, Engineering and Mathematics (STEM) > Mathematics and Statistics
Faculty of Science, Technology, Engineering and Mathematics (STEM)
|Depositing User:||Sarah Frain|
|Date Deposited:||16 Mar 2011 10:26|
|Last Modified:||02 Aug 2016 14:00|
|Share this page:|