Classifier dissimilarity-based

Dissimilarity-based classifiers. Dissimilarity-based classifiers (DBCs) use a dissimilarity measure Md to transform the input data into a dissimilarity space, where each trained class represents a separate dimension. Any dissimilarity measure with the following properties can be used ... 

Spatially resolved material identification and classification is currently the prevalent application for SI systems. Of the many powerful spectral classifiers available, only two types, each with a number of different algorithms,14 could successfully be applied for real-time SI applications discriminant classifiers and dissimilarity-based classifiers. In addition, occasionally dedicated algorithms, such as fuzzy-classifiers, may be useful for special applications, for example, when there is no ab inito knowledge about the number and properties of the classification classes. 

The idea is then to investigate different software metrics to see if metric values are similar within groups and dissimilar between groups. If this is the case, we can by calculating the metrics for a new randomly chosen program version, hopefully classify it. Based on the new program version s score results (which is available to us), we can evaluate our approach. Have we classified the program version correctly based on its metric values ... 

Inbred strains could be chosen on the basis of availability, knowledge of strain characteristics and absence of any biological properties which would preclude use of the strain, such as autoimmune disease or cancer. Strains which are genetically dissimilar could be chosen in order to maximise the chance of choosing at least one susceptible strain. For example inbred mouse strains have been classified into seven major families, based on the analysis... 

The efiectiveness of the BNB and NBN measures was assessed by simulated property-prediction experiments. These experiments involved the QSAR data sets studied previously by Pepperrell and Willetti" for the evaluation of distance-based similarity measures and a large set of 6-deoxyhexopyranose carbohydrates, which had previously been classified into 14 shape classes using numerical clustering methods based on torsional dissimilarity coefficients. The comparison encompassed the Bemis-Kuntz and Lederle measures, including not just the atom-triplet but also the atom-pair and atom-quadruplet versions of the former measure. The results were equivocal, in that it was impossible to... 

---