Abstract: We discuss the task of interpolating discrete probability densities resulting from the truncation of the binary representation of continuous random variables. We consider both the case in which the initial density is known and a simulator is to be programmed, and the case in which a random sample is given and a bootstrap resampling procedure is to be developed. We point out that the parametrisation of the discrete density suggested by the pRAM (probabilistic Random Access Memory) architecture and the use of Bernstein polynomials for its interpolation constitute an excellent hardware-software combination, able to provide a sagacious smoothing of the histogram of the truncated data.

Key-Words: pRAM, Bernstein polynomial, density estimation, bootstrap.
CSCC'99 Proceedings:-Pages 5361-5366