Showing papers on "Feature hashing published in 1993"

A new hashing function: statistical behaviour and algorithm

Abstract: Existing hashing functions have various limitations. In this paper a new hashing function is proposed, which divides the range of the key-values into some equal segments, and maps the key-values in each segment linearly into the whole range of the address. The paper analyzes the statistical behavior of the function, and points out that, theoretically, by increasing the number of segments, the distribution of the resulting hash values can always approach uniform, if the key-values can be regarded as continuous. Two methods for obtaining the number of segments, the deterministic and the probabilistic, along with the algorithm, are also proposed.

Differential hashing functions: application to reachability graph generation

Abstract: The paper presents the differential hashing functions. A differential computation process enables hashing function processing time to be optimized. After a formal definition of the differential property for hashing functions, we show that not all hashing functions have this property, then we propose a characterization of the differential hashing function set. Next, we show the performance acceleration produced by the differential algorithms applied to five hashing functions found in common applications. The observed accelerations can be significant because they are proportional to key length. Last we study the performances of a differential hashing function for the reachability graph exploration of distributed systems specified by a Petri net. This application demonstrates the advantages and the limits of the differential technique. >