[1Q3-J-2] Machine learning: structural modeling

[1Q3-J-2-01] L1 Regularization based Learning Method for Capsule Network

〇Nozomu Ohta¹, Shin Kawai¹, Hajime Nobuhara¹ (1. University of Tsukuba)

Keywords:Capsule Network

Capsule Network is a new neural network proposed to overcome the shortcomings of CNN. However, the Capsule
Network has many learnable parameters and is prone to over- tting. In this research, we aim to improve general-
ization ability by reducing parameters using L1 regularization. We evaluate our method by comparing the accuracy
and the reconstructed image with the conventional method.