An Information-Theoretic Analysis of Relational Databases—Part I: Data Dependencies and Information Metric

Answering Imprecise Queries over Autonomous Web Databases

“…Several authors [14,13,5] have proposed various measures to approximate the functional dependencies and keys that hold in a database. Among them, the g 3 measure proposed by Kivinen and Mannila [13], is widely accepted.…”

On Information-Theoretic Measures of Attribute Importance

“…If the functional dependency X → Y holds, we conclude that P (y j |x i ) is either 1 or 0. In term of conditional entropy, X → Y holds if and only if H(Y |X) = 0 [10,14]. By the relationships between entropy, conditional entropy, and mutual information, the above condition can be equivalently stated as H(X) = H(X, Y ) or I(X; Y ) = H(Y ) [10].…”

“…In term of conditional entropy, X → Y holds if and only if H(Y |X) = 0 [10,14]. By the relationships between entropy, conditional entropy, and mutual information, the above condition can be equivalently stated as H(X) = H(X, Y ) or I(X; Y ) = H(Y ) [10]. If Y is dependent on X, the partition of the database by X and Y is exactly the same as the one produced by X alone.…”

Why is the snowflake schema a good data warehouse design?

“…We also examine an information-theoretic interpretation of the snowflake schema following the work of [Mal86,CP87,Lee87,Mal88], which allows us to accommodate probabilistic information in the data warehouse. This is especially important, since decision making often involves probabilistic reasoning [Lin85].…”