Open Access. Powered by Scholars. Published by Universities.®
Physical Sciences and Mathematics Commons™
Open Access. Powered by Scholars. Published by Universities.®
Articles 1 - 2 of 2
Full-Text Articles in Physical Sciences and Mathematics
Names And Faces In The News, Tamara Berg, Alexander Berg, Jaety Edwards, Michael Maire, Ryan White, Yee Teh, Erik Learned-Miller, D Forsyth
Names And Faces In The News, Tamara Berg, Alexander Berg, Jaety Edwards, Michael Maire, Ryan White, Yee Teh, Erik Learned-Miller, D Forsyth
Erik G Learned-Miller
We show quite good face clustering is possible for a dataset of inaccurately and ambiguously labelled face images. Our dataset is 44,773 face images, obtained by applying a face finder to approximately half a million captioned news images. This dataset is more realistic than usual face recognition datasets, because it contains faces captured "in the wild" in a variety of configurations with respect to the camera, taking a variety of expressions, and under illumination of widely varying color. Each face image is associated with a set of names, automatically extracted from the associated caption. Many, but not all such sets …
Hyperspacings And The Estimation Of Information Theoretic Quantities, Erik G. Learned-Miller
Hyperspacings And The Estimation Of Information Theoretic Quantities, Erik G. Learned-Miller
Erik G Learned-Miller
The estimation of probability densities from data is widely used as an intermediate step in the estimation of entropy, Kullback-Leibler (KL) divergence, and mutual information, and for statistical tasks such as hypothesis testing. We propose an alternative to density estimation– partitioning a space into regions whose approximate probability mass is known–that can be used for the same purposes. We call these regions hyperspacings, a generalization of spacings in one dimension. After discussing one-dimensional spacings estimates of entropy and KL-divergence, we show how hyperspacings can be used to estimate these quantities (and mutual information) in higher dimensions. Our approach outperforms certain …