［ポスター講演］Direct Estimation of the Derivative of Quadratic Mutual Information with Application in Sufficient Dimension Reduction
○Voot Tangkaratt（Tokyo Inst. of Tech.）・Hiroaki Sasaki・Masashi Sugiyama（Univ. of Tokyo）
||(事前公開アブストラクト) In this paper we propose a novel method to directly estimate derivatives of Quadratic Mutual Information (QMI) without estimating the QMI itself.
The key idea is that, directly estimating derivative of a function, is more accurate than taken derivatives of an estimated function, which tends to produce a fluctuated derivatives. The result of the proposed method in supervised dimension reduction scheme is shown through artificial experiments.
||An accurate estimator of a function does not necessary mean that its derivative is an accurate estimator of the derivative of the function.
Motivated by this fact, we propose a method to directly estimate
the derivative of quadratic mutual information (QMI) without estimating QMI itself. QMI is a robust and stable variant of ordinary MI and is useful in various statistical data analysis tasks. We apply the proposed direct QMI derivative estimator to sufficient dimension reduction, and develop a natural gradient algorithm over the Grassmann manifold to find the most informative features. Finally, the usefulness of the proposed method is demonstrated through experiments.
|| / / / / / / /
||Mutual information / Dimensionality reduction / / / / / /
||信学技報, vol. 114, no. 306, IBISML2014-79, pp. 329-336, 2014年11月.
||Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380