A Novel Feature Extraction Algorithm Based on Joint Learning
Jeng‐Shyang Pan, Lijun Yan, Zongguang Fang · 2013
In this paper, a novel feature extraction algorithm, called Joint Discriminant Sparse Neighborhood Preserving Embedding (JDSNPE), based on Discriminant Sparse Neighborhood Preserving Embedding (DSNPE) and joint learning is proposed. JDSNPE aims to get the row sparsity of the transformation matrix while preserving discriminant sparse neighborhood. Experimental results on Yale database demonstrate the effectiveness of the proposed algorithm compared to Sparse Neighborhood Preserving Embedding and DSNPE.