Dimensionality Reduction using Relative Attributes

Abstract

Visual attributes are high-level semantic description of visual data that are close to the language of human. They have been intensively used in various applications such as image classification, active learning, and interactive search. However, the usage of attributes in dimensionality reduction has not been considered yet. In this work, we propose to utilize relative attributes as semantic cues in dimensionality reduction. To this end, we employ Non-negative Matrix Factorization (NMF) constrained by embedded relative attributes to come up with a new algorithm for dimensionality reduction, namely attribute regularized NMF (ANMF).