Lawrence Livermore National Laboratory, Livermore, CA
Pages: 11 Published: Jan 1987
So much can be done in infrared (IR) spectroscopy with computers that we are beginning to lose the healthy paranoia that should accompany their use. Even when we are careful, our precautions are quite unequal to the sophisticated and unobvious errors computers can generate. Even such apparently unobjectionable statements as “this algorithm should only be used by experts” often can be translated upon reflection to “since the program will track your prejudices, you better start with the right ones.” Searching algorithms, trial and error optimizations, multilinear regressions, and factor analysis all have higher order failure modes that are briefly reviewed. Computer simulation requires far more brute force in its application than we normally employ. And the very severe problem of flow frequency bugs can be approached only by redundant algorithmic pathways.
infrared, computers, algorithm, spectroscopy, computerized IR, Fourier deconvolution
Paper ID: STP23168S