Appeal No. 95-4609
Application 08/263,903
This is a decision on the appeal under 35 U.S.C. § 134
from the examiner's rejection of claims 1-7, which constitute all
the claims in the application.
The disclosed invention pertains to a method and
apparatus for predicting errors or failures in a communications
network. Measurements of failures during different, non-uniform
length intervals are determined by averaging the number of
failures during multiple uniform intervals. A neural network is
used to predict errors or failures based on these measurements.
Representative claim 1 is reproduced as follows:
1. A method for predicting errors or failures (i.e.,
“non-performances”) in a system for transmitting information
across a communications channel in accordance with non-
performances measured during periodic intervals, comprising the
steps of:
determining the number of non-performances occurring
during each of a plurality of non-uniform length, non-performance
measurement intervals by computing the average of the non-
performances over multiple uniform intervals;
selectively weighting the non-performance number for each
non-uniform interval;
summing the selectively weighted non-performance numbers
for the non-uniform intervals and applying a sigmoid function to
yield a set of intermediate values;
selectively weighting the intermediate values;
summing the selectively weighted intermediate values and
applying a sigmoid function to yield a set of predicted values
and;
2
Page: Previous 1 2 3 4 5 6 7 8 9 10 11 12 Next
Last modified: November 3, 2007