A sparse embedding and least variance encoding approach to hashing.

Area of Science:

• Computer Science
• Machine Learning
• Data Science

Background:

• Hashing is crucial for efficient large-scale image retrieval and data storage.
• Preserving kNN graphs in low-dimensional spaces is challenging with large datasets.

Purpose of the Study:

• To develop an effective and efficient hashing approach for large-scale image retrieval.
• To address the limitations of existing methods in handling big data.

Main Methods:

• Proposes sparse embedding of samples in the training space.
• Utilizes linear spectral clustering to partition the sample space.
• Employs a least variance encoding model with a learned dictionary for hash code generation.

Main Results:

• Experimental results on benchmark datasets demonstrate the effectiveness of the proposed approach.
• The method shows improved performance compared to state-of-the-art hashing techniques.

Conclusions:

• The proposed hashing approach is effective and efficient for large-scale image retrieval.
• Sparse embedding and dictionary learning offer a promising direction for hashing methods.