Learning Algorithms In The Infrared

Tomas Hirschfeld

doi:10.1117/12.970824

20 December 1985 Learning Algorithms In The Infrared

Tomas Hirschfeld

Proceedings Volume 0553, Fourier and Computerized Infrared Spectroscopy; (1985) https://doi.org/10.1117/12.970824
Event: 1985 International Conference on Fourier and Computerized Infrared Spectroscopy, 1985, Ottawa, Canada

Abstract

The use of learning algorithms to automatically set up analytical procedures in the near IR leads to considerable statistical complexities when one attempts to refine and optimize the method. While current work employs an almost brute force approach using lots of signal/noise ratios, plenty of samples and well defined representative learning sample sets, backed up by plenty of trial and error research, vastly more efficient, more easily automated, and more reliable procedures are, in principle, possible. It soon becomes apparent, however, that the power of our algorithms and the calculating capabilities of our computers are nowhere matched by our understanding of the underlying logic of the procedure. The learning algorithms of correlation spectroscopy, and the optimization process of multilinear regressions underlying them, are logically very complex entities whose complete understanding hinge upon a number of (seemingly) philosophical fine points. While no such complete understanding seems near at present (despite the overwhelming success of our so far crude approach to this spectroscopic technique), a number of insights derived from detailed considerations of the process will be discussed. These include procedures for avoiding optimization pitfalls, overall perform-ance improvements exploiting the non-Gaussian error distribution of the results, simplified optimization routines, and procedures for internal data verification.

Citation Download Citation

Tomas Hirschfeld "Learning Algorithms In The Infrared", Proc. SPIE 0553, Fourier and Computerized Infrared Spectroscopy, (20 December 1985); https://doi.org/10.1117/12.970824

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available