A method of generating random numbers

Abstract

The invention discloses a method for generating a random number, belonging to the computer field. The method for generating the random number is an improved algorithm of the interval algorithm under the condition of finite precision integer operation, and the main improvements include an implicit partition method of the output interval and a renormalization method of three sub-intervals of the output interval. This method completely eliminates the rounding error problem of the interval division of the output of the interval algorithm, but also greatly alleviates the rounding error problem of input interval partition, so that the problem that the quality of the output random number of the interval algorithm decreases due to the finite precision operation is basically solved, and the input data of any given probability distribution is converted into the random number of any given probability distribution in an accurate or near-accurate manner under the condition of the finite precision integer operation.

Description

technical field [0001] The invention relates to the field of computers, in particular to a method for generating random numbers. Background technique [0002] Random numbers are widely used in fields such as computer simulation and information security. The quality of random numbers, that is, the statistical distance between the real distribution of random numbers and the target distribution, is an important indicator for evaluating random number generation methods. Under the condition that the statistical model of the information source is known, the Interval Algorithm proposed in [1] can not only generate high-quality random numbers, but also has the highest efficiency in generating random numbers among various known methods. However, the original interval algorithm is based on infinite-precision arithmetic operations. When using limited-precision arithmetic operations, especially limited-precision integer operations, there are the following problems: [0003] 1. The pro...

AI Technical Summary

Problems solved by technology

[0003] 1. The rounding error problem of interval segmentation: division operation is required to divide the interval. In general, the division operation with limited precision will inevitably lead to rounding error

Improving the accuracy of operations is a common way to alleviate this problem, but it cannot eliminate it

[0004] 2. The problem that the input interval crosses the boundary of the output subinterval: During the operation of the interval algorithm, there will be situations where the input interval spans two adjacent output subintervals when the length is very small, so the sign of the next output cannot be determined, and the output interval cannot be valid As a result, the representation accuracy of the input ...

Images