Feature selection comparison

Tab. 5-13. Results of reclassification as a function of different feature sets comparison of the forward and the backward selection strategy...

Comparison of methods with the same type of feature selection forced on all methods, rather than letting each method do what it does best. 

In the following sections, selected comparisons between experimental kinetics for single-step electrochemical reactions and the foregoing electron-transfer models will be presented in order to characterize the physical features of the experimental systems as well as to scrutinize the applicability of the theoretical models themselves. 

The concept of parity in scaling up or down implies that the performance features, selectivity behavior, and recyclability of the stationary phase material used for the analytical and the preparative separation are identical, with the exception of particle size. Both robust experimental methods as well as rules of thumb, acquired by experienced investigators, have been developed that enable such comparisons to be made. An extensive scientific literature is now available to indicate sound foundations for such scaling up strategies, coupled with suitable experimental methods for their validation. Table 1 summarizes some of this information. 

Jeffery, I. B., Higgins, D. G., Culhane, A. C. 2006. Comparison and evaluation of microarray feature selection methods. In preparation. 

Table 1. Spectral features selected for the comparison of mass spectra...

Kupinski M, Giger ML (1999) Feature selection with limited datasets. Med Phys 26 2176-2182 Leichter I, Fields S, Nirel R, et al (2000) Improved mammographic interpretation of masses using computer-aided diagnosis. Eur Radiol 10 377-383 Loo L-N, Doi K, Metz CE (1984) A comparison of physical image quality indices and observer performance in the radiographic detection of nylon beads. Phys Med Biol 29 837-856... 

The fact that mass, infra-red, NMR and ultra-violet spectra and, to a certain extent, also chromatographic retention data can be considered as molecular fingerprints , forms the basis of most computerised library search systems. Retrieval methods for characteristic chemical data and techniques for the comparison of human fingerprints have similar elements the first step is to clean up the raw data, then in many cases a data reduction is carried out by selection of prominent features. Finally, there is the comparison of unknown and reference data patterns, which, for a useful result, requires a statistical correlation to be established. In this paper no attention will be paid to feature selection. 

Kleppa and his coworkers measured the enthalpies of formation of a series of intermetallic compounds between rare earth metals and transition metals by accurate high temperature calorimetric method [62]. Based on these data, we have used SVR-based feature selection method (this method is described in Chapter 4 of this book) to search the optimal set of atomic parameters for support vector regression. It has been found that an effective feature set suitable for the computerized prediction of the enthalpy of formation of these compounds includes only two parameters Miedema s A< ) and the ratio of atomic radii R1/R2 (here the values of the atomic radius proposed by Teatum are used). By SVR and leave-one-out (LOO) cross-validation, the results of prediction are rather good. Table 7.1 shows the comparison between the experimental values and the predicted values of the enthalpy of formation of these compounds. The results of calculation by Miedema s model [104] are also listed for comparison. 

Lorente, D. et al. (2013) Selection of optimal wavelength features for decay detection in citrus fruit using the ROC curve and neural networks. Food Bioprocess TechnoL, 6 (2), 530-541. Lorente, D. et al (2012) Comparison of ROC feature selection method for the detection of decay in citrus fruit using hyperspectral images. Food Bioprocess TechnoL, 6 3613-3619. 

Five methods of feature selection (information gain, mutual information, X -test, odds ratio, and GSS coefficient) were compared by Liu for their ability to discriminate between thrombin inhibitors and noninhibitors.The chemical compounds were provided by DuPont Pharmaceutical Research Laboratories as a learning set of 1909 compounds contained 42 inhibitors and 1867 noninhibitors, and a test set of 634 compounds contained 150 inhibitors and 484 noninhibitors. Each compound was characterized by 139,351 binary features describing their 3-D structure. In this comparison of naive Bayesian and SVM classifiers, all compounds were considered together, and a L10%O cross-validation procedure was applied. Based on information gain descriptor selection,... 

The analysis of cigarette smoke for 16 different polyaromatic hydrocarbons is described in this experiment. Separations are carried out using a polymeric bonded silica column with a mobile phase of 50% v/v water, 40% v/v acetonitrile, and 10% v/v tetrahydrofuran. A notable feature of this experiment is the evaluation of two means of detection. The ability to improve sensitivity by selecting the optimum excitation and emission wavelengths when using a fluorescence detector is demonstrated. A comparison of fluorescence detection with absorbance detection shows that better detection limits are obtained when using fluorescence. 

Whenever an economic evaluation is undertaken, a corresponding problem definition should be provided as the basis on which the evaluation is made. This definition, sometimes called an economic scope, should clearly differentiate between specifications that have actually been selected and features that have been assumed for the evaluation. In a comparison of alternatives, all of the assumptions, data, and conditions must be consistent, reaUstic, and devoid of bias. 

The general construction features and materials are quite similar to the lohe type units. For comparison selections, refer to Huff, Patton,Price,Ahraham, and Van Ormer. ... 

Thin sections cut with a diamond knife microtome can be of great advantage in locating regions of catalyst where important chemical or structural changes take place during reaction. Comparison of equivalent areas of fresh and deactivated catalyst can be a difficult problem if the catalyst support does not have a uniform microstructure as in carbon supports produced from plant materials. Even when specimen selection and preparation are adequate, it may be difficult to know upon which image features to place the electron beam to solve the problem at hand. 

Micellar flooding is a promising tertiary oil-recovery method, perhaps the only method that has been shown to be successful in the field for depleted light oil reservoirs. As a tertiary recovery method, the micellar flooding process has desirable features of several chemical methods (e.g., miscible-type displacement) and is less susceptible to some of the drawbacks of chemical methods, such as adsorption. It has been shown that a suitable preflush can considerably curtail the surfactant loss to the rock matrix. In addition, the use of multiple micellar solutions, selected on the basis of phase behavior, can increase oil recovery with respect to the amount of surfactant, in comparison with a single solution. Laboratory tests showed that oil recovery-to-slug volume ratios as high as 15 can be achieved [439]. 

---