학술논문
Code generator matrices as RNG conditioners.
Document Type
Journal
Author
Tomasi, A. (I-TRNT) AMS Author Profile; Meneghetti, A. (I-TRNT) AMS Author Profile; Sala, M. (I-TRNT) AMS Author Profile
Source
Subject
11 Number theory -- 11T Finite fields and commutative rings
11T71Algebraic coding theory; cryptography
60Probability theory and stochastic processes -- 60B Probability theory on algebraic and topological structures
60B99None of the above, but in this section
94Information and communication, circuits -- 94B Theory of error-correcting codes and error-detecting codes
94B99None of the above, but in this section
11T71
60
60B99
94
94B99
Language
English
Abstract
Let $ X \in (\Bbb F_p)^n $ be a vector with independent random coordinates taking values in a finite field. The paper under review derives a bound on the total variation distance of $ X $ from the uniform distribution in the binary case $ p=2 $, expressed in terms of the sum of biases of individual bits. This bound is then generalized to the product $ GX $ in place of $ X $, where $ G $ is the generator matrix of a linear code over $ \Bbb F_p = \Bbb F_2 $, and subsequently extended to the case of a general $ p $. A unified treatment is given, based on the Walsh-Hadamard transform and the number theoretic transform.