Library compression

There are few possible strategies of library compression. Each of them has its own advantages and drawbacks. The most efficient method of data set compression, i.e. Principal Component Analysis (PCA), leads to use of global features. As demonstrated in [15] global features such as PCs (or Fourier coefficients) are not best suited for a calibration or classification purposes. Often, quite small, well-localized differences between objects determine the very possibility of their proper classification. For this reason wavelet transforms seem to be promising tools for compression of data sets which are meant to be further processed. However, even if we limit ourselves only to wavelet transforms, still the problem of an approach optimally selected for a particular purpose remains. There is no single method, which fulfills all requirements associated with a spectral library s compression at once. Here we present comparison of different methods in a systematic way. The approaches A1-A4 above were applied to library compression using 21 filters (9 filters from the Daubechies family, 5 Coiflets and 7 Symmlets, denoted, respectively as filters Nos. 2-10, 11-15 and 16-22). 

Table 1. Results of the library compression using iDWT and iWPT approaches, and MDL and RMS criteria mean, minimal (min), and maximal (max) values, and standard deviation (std) of Root Mean Squares Error (RMS), and the number of the retained coefficients (N). 

Table 2. Results of the library compression using JB99 and JBB99 approaches filter No. 15.

The IR library, compressed according to the discussed approaches, was tested for performance in matching. For this purpose, the spectra with 1 % and 5% of random white noise added (see Fig. 13) were decomposed and represented in the same basis as the compressed library, and then each of them was matched with all spectra from the compressed library. The same operation was performed for the data set in the original domain. 

These results encouraged us to further reduce the number of wavelet coefficients used for spectra matching. To this end, the matching performance using different number of wavelet coefficients has been studied, for library compressed with JBB approach. The results are presented in Fig. 14. 

Obviously, no single approach fulfills all requirements associated with library compression. Taking into account all aspects of the library compression and searching, we have to find a compromise among all different requirements. It means we should base our final choice of the optimal strategy on criteria described above, weighting their respective importance. 

Most mass spectrometers for analytical work have access to a large library of mass spectra of known compounds. These libraries are in a form that can be read immediately by a computer viz., the data corresponding to each spectrum have been compressed into digital form and stored permanently in memory. Each spectrum is stored as a list of m/z values for all peaks that are at least 5% of the height of the largest peak. To speed the search process, a much shorter version of the spectrum is normally examined (e.g., only one peak in every fourteen mass units). 

There will be interesting features within some of the components models that not all of them can deal with. Not all televisions can deal with color signals not all fax machines understand Group in compression not all word processors understand tables not all of the software components running a library will understand the concept of the acquisition date of a book even though most of them will understand its title. 

Problem formulations [ 1-3 ] for designing lead-generation library under different constraints belong to a class of combinatorial resource allocation problems, which have been widely studied. They arise in many different applications such as minimum distortion problems in data compression (11), facility location problems (12), optimal quadrature rules and discretization of partial differential equations (13), locational optimization problems in control theory (9), pattern recognition (14), and neural networks... 

Moreover, we have included a zip file CBE-book.zip on the CD with the contents of the CD itself in compressed form. This zip file is used for easy installation of our m file library on any hard drive. Local installation of the m files will facilitate handling of our codes under MATLAB. 

In your university library, find the paper Barone, M.R. and T.A. Osswald, J. of Non-Newt. FluidMech., 26,185-206, (1987), and write a 2D FEM program to simulate the compression molding process using the Barone-Caulk model presented in the paper. Compare your results to the BEM results presented in the paper. 

It is not necessary for the physicist to know how to compute the Coulomb functions. They are found in subroutine libraries, for example Barnett et al. (1974). A sufficient idea of their form is obtained by putting j = L = 0 in (4.62), when they are seen to be sinp and cosp respectively. The potential terms dilate or compress the sine and cosine waves, resulting in an overall phase shift at long range. 

First, the chapter lists the possible unit operations in the Aspen Plus Model Library, because the process is a connected set of the units. Then an example process is illustrated that makes ammonia from nitrogen and hydrogen. You will be able to get both the mass balances and the energy balances for the process. With this information you can determine the size of most of the equipment needed, and hence its cost. You can also determine the operating cost for heating, cooling, compression, and other tasks. The process involves a... 

When several libraries are mapped into one SOM they must compete for the space on the map. The most diverse library occupies most of the neurons. Less diverse libraries are compressed to a small region on the map. 

In the present chapter, we apply an accurate and numerically efficient equation of state for the exp-6 fluid based on Zerah and Hansen s hypemetted-mean spherical approximation (HMSA)[22] equations and Monte Carlo calculations to detonation, shocks, and static compression. We present a library of parameters for fluid and condensed high pressure molecules in Ref... 

Figure 1.15 Automated sorting facility of plastic wastes based on IR radiation27 IR source (1), IR detector (2), computer with spectrum library (3), main conveyor (4), compressed air (5), solenoid valves (6), pusher (7).

For those deeply concerned about perimeter defense of their property, there is the Queen of Battle, artillery. A compressed air powered cannon is easily constructed. Building them is so simple that I ve seen them featured on the Junkyard Wars TV show. The bore size of the cannon can be chosen so that the common aluminum soda pop can will fit perfectly. This can, once loaded with an explosive charge, makes an efficient projectile. Air cannons easily propel a loaded soda can 100 yards or more. For details on building an air cannon, see US Patent 4,703,869. Your local library will gladly send for a copy of the patent upon your request. You can also read the body of the patent minus the drawings by going to the US patent office website, and typing in the patent number. 

WT has been proposed as a new method for compressing spectra for storage and library searching in our study. In this kind of work, spectra are reconstructed from time to time from the compressed data. In order to maintain the quality of the reconstructed spectra, we have introduced another technique called the translation-rotation transformation (TRT) method [23] in the wavelet computation. In the FWT operation, the spectral data vector Cj needs to be extended periodically at the two extremes in the following manner ... 

A.K.M. Leung, F.T. Chau. J.B. Gao and T.M. Shih. Application of Wavelet Transform in Infrared Spectrometry Spectral Compression and Library Search, Chcmometric Intelligent Laboratory System. 43 (1998). 69-88. 

Despite recent advances, the problem of spectral libraries size build-up, and search speed receives still considerable amount of attention. Most of the commercial databases to date use Fast Fourier Transform (FFT) for spectra compression. However, the past ten years have brought explosive growth of wavelet applications in signal processing. The IR spectra show many ab-... 

The speed of matching for a compressed library depends on the number N of retained coefficients. In the case of individual compression, the average number of wavelet coefficients is taken into account. Compression ratio can be calculated as ... 

For discussion of the storage requirements of applying wavelet methods to compress spectral libraries we ought to consider the following situation ... 

A normalized scalar product of two spectral vectors was used during matching as the similarity measure, and sequential searches through the entire library were always performed. Thus, each spectrum in turn was treated as a query. To simulate small variances in data acquisition, and/or spectral differences for very similar compounds, 1% and 5% of random white noise, has been added. Appropriate decomposition (wavelet or PCA) was then performed, and the resulting vector was compared to each of the spectral vectors in the compressed library. 

The most efficient method of data set compression in the joint basis is Principal Component Analysis (PCA). Principal Components (PCs) are constructed as a linear combination of original variables to maximize the description of data variance. They are eigenvectors of the auto-covariance matrix of data set. Each eigenvector is associated with the corresponding eigenvalue, which describes its importance in data variance description. For the studied IR library, 57 eigenvectors (principal components) are necessary to describe 95% of data variance, whereas as much as 109 eigenvectors are needed to describe 99% of data variance (see Fig.5). The mean value of RMS... 

---