Browse Prior Art Database

Adaptive Model for Nonstationary Sources

IP.com Disclosure Number: IPCOM000061416D
Original Publication Date: 1986-Jun-01
Included in the Prior Art Database: 2005-Mar-09
Document File: 1 page(s) / 12K

Publishing Venue

IBM

Related People

Rissanen, J: AUTHOR [+2]

Abstract

A method is described for adaptively estimating the probability of occurrence of a symbol from a binary nonstationary source. The method comprises the steps of (1) assigning an initial value to said probability; and (2) updating said probability value at the occurrence of each symbol, the updating being dependent on a combination of the occurred symbol and a Monte Carlo function.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 85% of the total text.

Page 1 of 1

Adaptive Model for Nonstationary Sources

A method is described for adaptively estimating the probability of occurrence of a symbol from a binary nonstationary source. The method comprises the steps of (1) assigning an initial value to said probability; and (2) updating said probability value at the occurrence of each symbol, the updating being dependent on a combination of the occurred symbol and a Monte Carlo function.

Let qk(s) denote the probability of occurrence of the least probable symbol (LPS) of a binary nonstationary source after a string s of symbols is observed. With little loss in performance, the probability of occurrence of the LPS can be restri to a finite set [Q1(S),Q2(S),...,QK(S)] where q1(s) = 1/2 and qi(s) > q(i+1)(s). If the next symbol x observed after s is an LPS, then the assumed value of qk(s) may need to be increased. However, the occurrence of an LPS in x also depends on a Monte Carlo function. Therefore, q should be increased if the next symbol x observed is an LPS and if qk(s) is equal to or less than a random variable y having a uniform distribution over the interval [0,1]. Conversely, qk(s) should be decreased if the next symbol x observed is a most probable symbol (MPS) and if qk(s) is also greater than a Monte Carlo generated random variable y having a uniform distribution over the interval [0,1]. In general, the probability of occurrence of the LPS is given by the following: qk(sx) = q(k- 1)(s) if x = LPS and y < qk(s) _ q(k+1)(s...