Geometrical Perspective on Learning Behavior
Cheng-Yuan Liou, Jau-Chi Huang and Yen-Ting Kuo Department of Computer Science and Information Engineering National Taiwan University Taipei, 106 Taiwan E-mail: cyliou@csie.ntu.edu.tw
This paper constructs a geometrical perspective to justify the slow learning period and fast learning period during training. It plots the error surfaces and the solution spaces in the input space for a single neuron with two inputs. It records various training paths in this space using the back-propagation (BP) training algorithm [6]. It finds the relations between the learning curve and training path.