Heterogeneous data

For homogeneous NDT data and repeatable inspection conditions successful automated interpretation systems can relatively easily be developed. They usually use standard techniques from statistical classification or artificial intelligence. Design of successful automated interpretation systems for heterogeneous data coming form non-repeatable, small volume inspections with little a-priori information about the pieces or constructions to be inspected is far more difficult. This paper presents an approach which can be used to develop such systems. 

First, the typical characteristics of inspection problems which result in heterogeneous data are presented. Next, typical AI techniques which can be used for the automated data interpretation are presented. The applicabihty of the techniques to various inspection problems is discussed. Two example apphcations for automatic NDT data interpretation are briefly described, and finally, the conclusions are given. 

The second task discussed is the validation of the regression models with the aid of the cross-validation (CV) procedures. The leave-one-out (LOO) as well as the leave-many-out CV methods are used to evaluate the prognostic possibilities of QSAR. In the case of noisy and/or heterogeneous data the LM method is shown to exceed sufficiently the LS one with respect to the suitability of the regression models built. The especially noticeable distinctions between the LS and LM methods are demonstrated with the use of the LOO CV criterion. 

Three examples concerning toxicity of heterogeneous data sets are given below. The first [50] relates to mutagenicity to Salmonella typhimurium of aromatic and heteroaromatic nitro-compounds ... 

To extract this knowledge from the various heterogeneous data sources made accessible to the KSP and the UltraLink, we combine several steps of normalization and analysis. These procedures are applied at loading time whenever the documents are displayed in the browser. 

TABLE 5.12. Sample Data Sets Homogeneous versus Heterogeneous Data... 

Accommodate access to heterogeneous data sources and analysis tools. 

Atkinson R., Baulch D.L., Cox R.A., Hampson Jr. R.F., Kerr J. A., Rossi M.J., and Troe J. (1997) Evaluated kinetic, photochemical and heterogeneous data for atmospheric chemistry. Supplement V, lUPAG Subcommittee on Gas Kinetic Data Evaluation for Atmospheric Chemistry. /. Phys. Chem. Ref. Data 26, 521-1011. 

Atkinson, R D. L. Baulch, R. A. Cox, R. F. Hampson, Jr J. A. Kerr, M. J. Rossi, and J. Troe, Evaluated Kinetic and Photochemical Data for Atmospheric Chemistry Supplement VI. IUPAC Subcommittee on Gas Kinetic Data Evaluation for Atmospheric Chemistry, J. Phys. Chem. Ref. Data, 26, 1329-1499 (1997a). Atkinson, R., D. L. Baulch, R. A. Cox, R. F. Hampson, Jr., J. A. Kerr, M. J. Rossi, and J. Troe, Evaluated Kinetic, Photochemical, and Heterogeneous Data for Atmospheric Chemistry Supplement V. IUPAC Subcommittee on Gas Kinetic Data Evaluation for Atmospheric Chemistry, J. Phys. Chem. Ref. Data, 26, 521-1011 (1997b). 

However, although this equation was effective in modelling the odour thresholds of the disubstituted pyrazines, two main weaknesses have been identified (72) the first was that it was difficult to dmw physical meaning from the descriptor AA J, since it was not clear which aspects of die molecular structure determined the odour threshold. The second we ess was discovered when pyrazine itself and thirteen mono-substituted pyrazines were added to the original set. The calculated and observed odour threshold values were no longer in agreement. This result indicated diat the model was insufficient for more heterogeneous data sets. 

In order to reduce the error from simply aggregating such diverse estimates, an attempt was made to standardize - as a far as possible - the very heterogeneous data set. Thus, all available estimates were transformed... 

Within a homologous series it may be better to rank relative water solubility (rather than relying on specific calculated values). Their use with heterogeneous data sets is even more fraught, and calculated values should be used with considerable circumspection. 

P. C. Jurs, J. T. Chou, and M. Yuan, J. Med. Chem., 22, 476 (1979). Computer-Assisted Structure—Activity Studies of Chemical Carcinogens. A Heterogeneous Data Set. 

In step 1, the initial compounds are usually selected at random from the data set. Random selection is quick and, for large heterogeneous data sets, likely to provide a reasonable initial set. Steps 2 and 3 can be performed separately or in combination. If done separately, the classification (step 2) is performed on... 

---