Abstract
In this paper, we propose a new simple estimate for the width of the Gaussian kernel, based on the estimate proposed by Nakayama. First, the simple estimate for width proposed by Nakayama is analyzed, and then some sufficient conditions will be described. According to these sufficient conditions, a new simple estimate for the width of the Gaussian kernel is proposed. Considering the equivalence between some machine learning techniques, it is expected that the proposed estimate of the width is applicable to wide range of machine learning techniques employing the Gaussian kernel. Through numerical examples, the validity of the proposed estimate of the width is examined.
