Effective Data-Aware Covariance Estimator From Compressed Data

Area of Science:

• Statistics
• Machine Learning
• Data Science

Background:

• Estimating covariance matrices is crucial for analyzing massive, high-dimensional, and distributed data.
• Existing methods face challenges in accuracy and efficiency with large-scale datasets.

Purpose of the Study:

• To propose a novel, data-aware weighted sampling-based covariance matrix estimator, named DACE.
• To enhance the accuracy and efficiency of covariance matrix estimation for massive datasets.
• To extend the DACE method for multiclass classification problems.

Main Methods:

• Developed a data-aware weighted sampling strategy for covariance estimation.
• The DACE algorithm provides an unbiased covariance matrix estimation.
• Extended DACE for multiclass classification with theoretical underpinnings.

Main Results:

• DACE achieves more accurate covariance matrix estimation under the same compression ratio compared to existing methods.
• Demonstrated superior performance of DACE on both synthetic and real-world datasets.
• Validated the effectiveness of the extended DACE for multiclass classification tasks.

Conclusions:

• DACE offers a significant advancement in estimating covariance matrices from massive, high-dimensional, and distributed data.
• The proposed method provides a robust and accurate solution for data analysis and machine learning applications.
• DACE shows promise for improving multiclass classification performance.