PROJECT TITLE :

Robust Joint Feature Weights Learning Framework

ABSTRACT:

Feature selection, choosing the most informative subset of options, is an important research direction in dimension reduction. The combinatorial search in feature choice is essentially a binary optimization drawback, known as NP laborious, that will be alleviated by learning feature weights. Ancient feature weights algorithms depend upon heuristic search path. These approaches neglect the interaction and dependency between different features, and thus give no guarantee for optimality. In this paper, we have a tendency to propose a completely unique joint feature weights learning framework, that imposes each nonnegative and -norm constraints on the feature weights matrix. The nonnegative property ensures the physical significance of learned feature weights. Meanwhile, -norm minimization achieves joint selection of the most relevant options by exploiting the whole feature space. More importantly, an efficient iterative algorithm with proved convergence is designed to optimize a convex objective operate. Using this framework as a platform, we tend to propose new supervised and unsupervised joint feature choice strategies. Significantly, within the proposed unsupervised method, nonnegative graph embedding is developed to use intrinsic structure in the weighted house. Comparative experiments on seven real-world knowledge sets indicate that our framework is each effective and efficient.