Comparing hard and soft prior bounds in geophysical inverse problems

Backus, George E.

In linear inversion of a finite-dimensional data vector y to estimate a finite-dimensional prediction vector z, prior information about X sub E is essential if y is to supply useful limits for z. The one exception occurs when all the prediction functionals are linear combinations of the data functionals. Two forms of prior information are compared: a soft bound on X sub E is a probability distribution p sub x on X which describeds the observer's opinion about where X sub E is likely to be in X; a hard bound on X sub E is an inequality Q sub x(X sub E, X sub E) is equal to or less than 1, where Q sub x is a positive definite quadratic form on X. A hard bound Q sub x can be softened to many different probability distributions p sub x, but all these p sub x's carry much new information about X sub E which is absent from Q sub x, and some information which contradicts Q sub x. Both stochastic inversion (SI) and Bayesian inference (BI) estimate z from y and a soft prior bound p sub x. If that probability distribution was obtained by softening a hard prior bound Q sub x, rather than by objective statistical inference independent of y, then p sub x contains so much unsupported new information absent from Q sub x that conclusions about z obtained with SI or BI would seen to be suspect.

Document ID

19880004438

Acquisition Source

Legacy CDMS

Document Type

Contractor Report (CR)

Authors

Date Acquired

September 5, 2013

Publication Date

December 28, 1987

Subject Category

Report/Patent Number

Accession Number

88N13820

Funding Number(s)

Distribution Limits

Public

Work of the US Gov. Public Use Permitted.

Available Downloads

Name

Type

19880004438.pdf

STI

No Preview Available

NTRS

NTRS - NASA Technical Reports Server

Available Downloads

Related Records