Efficient Techniques Based on Sparse Representation for Classifying High-dimensional Multiclass Microarray Data

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• High-dimensional data, such as microarray gene expression profiles, pose challenges for conventional classifiers due to their complexity and limited sample sizes.
• Sparse Representation (SR) offers a powerful alternative for classification tasks involving such data.
• SR methods can reduce computational complexity and enhance classification accuracy.

Purpose of the Study:

• To explore and evaluate Sparse Representation (SR)-based classifiers for microarray data classification.
• To investigate the impact of different dictionary construction strategies and sparse coding algorithms on classification performance.
• To address the computational cost associated with using all training samples in the SR dictionary.

Main Methods:

• The study explored various SR-based classifiers tailored for microarray data.
• Focus was placed on optimizing dictionary construction strategies, including selecting representative atoms.
• The SL0 algorithm was investigated as a sparse coding method.

Main Results:

• Experimental results were obtained using the 14-Tumors dataset.
• Selecting a subset of representative atoms significantly improved classification speed.
• Employing the SL0 algorithm further enhanced both classification speed and accuracy.

Conclusions:

• Sparse Representation (SR) approaches demonstrate significant potential for classifying high-dimensional biological data.
• Optimized dictionary construction and efficient sparse coding algorithms are crucial for effective SR-based classification.
• The findings suggest SR methods can provide efficient and accurate solutions for complex biological data analysis.