Geometric Mean of Partial Positive Definite Matrices with Missing Entries

Hayoung Choi, Sejong Kim, Yuanming Shi

Abstract: In this paper the geometric mean of partial positive definite matrices with missing entries is considered. The weighted geometric mean of two sets of positive matrices is defined, and we show whether such a geometric mean holds certain properties which the weighted geometric mean of two positive definite matrices satisfies. Additionally, counterexamples demonstrate that certain properties do not hold. A Loewner order on partial Hermitian matrices is also defined. The known results for the maximum determinant positive completion are developed with an integral representation, and the results are applied to the weighted geometric mean of two partial positive definite matrices with missing entries. Moreover, a relationship between two positive definite completions is established with respect to their determinants, showing relationship between their entropy for a zero-mean, multivariate Gaussian distribution. Computational results as well as one application are shown.

Introduction

In fields like statistics, signal processing, and image analysis, data is often represented as Positive Definite (PD) matrices, such as covariance matrices. The geometric mean is a powerful tool for averaging these matrices because it respects their underlying non-Euclidean geometry. However, real-world data is frequently incomplete due to sensor failure or unavailable measurements, resulting in "partial" matrices where some entries are missing. Standard techniques often involve deleting incomplete records, which wastes valuable information. This research addresses how to calculate a mathematically rigorous geometric mean for two matrices when not all values are known.

Graph Theory and Matrix Completion

The authors bridge the gap between matrix analysis and graph theory by using undirected graphs to represent the structure of a partial matrix. In these graphs, edges correspond to known entries, while missing edges represent unknown values.

A critical concept in this framework is "chordality." A graph is chordal if every cycle of length four or more has a "shortcut" or chord. If the graph of known entries is chordal, a positive definite completion is guaranteed to exist.

Comparison of a non-chordal cycle graph (left) and a chordal graph with an added edge (right).

Methods and Framework

Since a partial matrix represents a set of all possible valid completions, the authors define the geometric mean of two partial matrices as the set of geometric means of all their respective completions. To find a unique representative from these sets, they utilize the "Maximum Determinant" approach. This specific completion is significant because it corresponds to the maximum entropy of a multivariate Gaussian distribution—essentially filling in the blanks with the "least biased" information.

Additionally, the paper defines a Loewner order for incomplete Hermitian matrices. This allows for "greater than" or "less than" comparisons even in the presence of missing data, facilitating optimization in incomplete datasets.

Key Insights and Results

  • The Chordal Requirement: The paper reinforces that chordal graphs are the "gold standard" for matrix completion. Without chordality, a partial matrix may have no valid positive definite completion.
  • Entropy Connection: A major finding is the link between the geometric mean and Shannon entropy. The authors demonstrate that the difference in entropy between two Gaussian distributions can be expressed through an integral involving the geometric mean of their covariance matrices.
  • Property Breakdown: Interestingly, some properties of the standard geometric mean do not carry over to the partial case. For example, while $A \# A = A$ for full matrices, for partial matrices, the set-based mean of a partial matrix with itself is actually a superset of its original completions.
  • Uniqueness of Max-Det Mean: Theorem 6.11 establishes that the maximum determinant completion of the geometric mean of two partial matrices is exactly the geometric mean of their individual maximum determinant completions.

Conclusion

This work provides a robust framework for averaging incomplete covariance matrices by leveraging maximum determinant completions and graph theory. While individual completion sets are convex, determining if the resulting geometric mean set is always convex remains an open question. Future work may also address the computational complexity of implementing these methods for very large-scale matrices with non-chordal missing data patterns.

Summary updated

Abstract

In this paper the geometric mean of partial positive definite matrices with missing entries is considered. The weighted geometric mean of two sets of positive matrices is defined, and we show whether such a geometric mean holds certain properties which the weighted geometric mean of two positive definite matrices satisfies. Additionally, counterexamples demonstrate that certain properties do not hold. A Loewner order on partial Hermitian matrices is also defined. The known results for the maximum determinant positive completion are developed with an integral representation, and the results are applied to the weighted geometric mean of two partial positive definite matrices with missing entries. Moreover, a relationship between two positive definite completions is established with respect to their determinants, showing relationship between their entropy for a zero-mean,multivariate Gaussian distribution. Computational results as well as one application are shown.
BibTeX
@inproceedings{choi2019geometric,
  author = {Hayoung Choi and Sejong Kim and Yuanming Shi},
  title = {{Geometric Mean of Partial Positive Definite Matrices with Missing Entries}},
  booktitle = {Linear},
  year = {2019},
  doi = {10.1080/03081087.2019.1585741},
}
Figure