Method for guessing the response of a physical system to an arbitrary input

Patent · Mon Jan 01 00:00:00 EST 1996

OSTI ID:870508

Wolpert, David H ^[1]

---

1. Santa Fe, NM

Stacked generalization is used to minimize the generalization errors of one or more generalizers acting on a known set of input values and output values representing a physical manifestation and a transformation of that manifestation, e.g., hand-written characters to ASCII characters, spoken speech to computer command, etc. Stacked generalization acts to deduce the biases of the generalizer(s) with respect to a known learning set and then correct for those biases. This deduction proceeds by generalizing in a second space whose inputs are the guesses of the original generalizers when taught with part of the learning set and trying to guess the rest of it, and whose output is the correct guess. Stacked generalization can be used to combine multiple generalizers or to provide a correction to a guess from a single generalizer.

• Research Organization: Los Alamos National Laboratory (LANL), Los Alamos, NM (United States)
• Assignee: United States of America as represented by United States (Washington, DC)
• Patent Number(s): US 5535301
• OSTI ID: 870508
• Country of Publication: United States
• Language: English

Related Subjects

method
guessing
response
physical
arbitrary
input
stacked
generalization
minimize
errors
generalizers
acting
set
values
output
representing
manifestation
transformation
hand-written
characters
ascii
spoken
speech
computer
command
etc
deduce
biases
generalizer
respect
learning
correct
deduction
proceeds
generalizing
space
inputs
guesses
original
taught
trying
guess
combine
multiple
provide
correction
single
output values
input values
values representing
output value
/706/