Skip to Main Content
A high-dimensional database system is studied where the noisy versions of the underlying feature vectors are observed in both the enrollment and query phases. The noisy observations are compressed before being stored in the database, and the user wishes to both identify the correct entry corresponding to the noisy query vector and reconstruct the original feature vector within a desired distortion level. A fundamental capacity-storage-distortion tradeoff is identified for this system in the form of single-letter information theoretic expressions. The relation of this problem to the classical Wyner-Ziv rate-distortion problem is shown, where the noisy query vector acts as the correlated side information available only in the lossy reconstruction of the feature vector.