Title :
On Geometric Structure of Quasi-Additive Learning Algorithms
Author_Institution :
Kyoto Univ., Kyoto
Abstract :
Quasi-additive (QA) algorithms are a kind of online learning algorithms having two parameter vectors: one is an accumulation of input vectors and the other is a weight vector for prediction associated with the former by a non-linear function. We show that the vectors have a dually-flat structure from the information-geometric point of view, which makes it easier to discuss the convergence properties of the algorithms, as presented here.
Keywords :
geometry; learning (artificial intelligence); nonlinear functions; convergence properties; information-geometric structure; nonlinear function; parameter vectors; quasiadditive learning algorithms; weight vector; Algorithm design and analysis; Convergence; Inference algorithms; Information analysis; Information geometry; Information theory; Mathematical programming; Neural networks; Nonlinear equations; Physics;
Conference_Titel :
Neural Networks, 2006. IJCNN '06. International Joint Conference on
Conference_Location :
Vancouver, BC
Print_ISBN :
0-7803-9490-9
DOI :
10.1109/IJCNN.2006.246816
