Conceptual clustering

Gelemter and Rose [25] used machine learning techniques Chapter IX, Section 1.1 of the Handbook) to analyze the reaction center. Based on the functionalities attached to the reaction center, the method of conceptual clustering derived the features a reaction needed to possess for it to be assigned to a certain reaction type. A drawback of this approach was that it only used topological features, the functional groups at the reaction center, and its immediate environment, and did not consider the physicochemical effects which are so important for determining a reaction mechanism and thus a reaction type. 

Kohonen network Conceptual clustering Principal Component Analysis (PCA) Decision trees Partial Least Squares (PLS) Multiple Linear Regression (MLR) Counter-propagation networks Back-propagation networks Genetic algorithms (GA)... 

Michalski, R. S. Stepp, R. E. "Learning from Observation Conceptual Clustering" In "Machine Learning An Artificial Intelligence Approach" Michalski, R. S. Carbonell, J. G. Mitchell, T. M., Eds. Tioga Press, 1983. 

Symposium, pp. 91-95, 2000 (http //www-cse.uta.edu/ cook/pubs). Graph-Based Hierarchical Conceptual Clustering. 

Fisher, D. H. 1987. Knowledge acquistion via incremetal conceptual clustering. Mach. Learn. 2 139-172. 

Michalski, R.S. (1980). Knowledge acquisition through conceptual clustering A theoretical framework and algorithm for partitioning data into conjunctive concepts, Int. J. Policy Anal. Inf. Syst., 4, 219-243. 

Okada, T. Wipke, W. T. CLUSMOL A System for the Conceptual Clustering of Molecules . Tetrahedron Comput. Methodol. 9S%j 2, 249-264. 

Conceptual clustering is an extension of numerical taxonomic methods which uses domain knowledge to express the relationships between attributes explicitly. Conceptual clustering also differs from numerical taxonomy in that clustering quality is not simply a function of the objects in the clusters but also of the concepts that the resulting clusters express. ... 

In contrast to the numerical taxonomy quality function, the conceptual clustering quality function takes into account the utility of the concepts representing the clusters, as seen below, where the parameter CONCEPTS is the set domain of concepts which may be used to describe the clusters. In the context of conceptual clustering, a concept is a valid cluster description as specified within the cluster description language. Consequently, the set of possible concepts need not be explicitly enumerated. An important bias in this method is the use of the simplicity of a cluster description as a metric for the quality of clustering. 

Conceptual clustering has been used hierarchically to classify chemical reactions. The concepts expressed by the... 

There are only a few weU-documented examples of catalysis by metal clusters, and not many are to be expected as most metal clusters are fragile and fragment to give metal complexes or aggregate to give metal under reaction conditions (39). However, the metal carbonyl clusters are conceptually important because they form a bridge between catalysts commonly used in solution, ie, transition-metal complexes with single metal atoms, and catalysts commonly used on surfaces, ie, small metal particles or clusters. 

To sum up, theory plays an important role in the understanding of cluster properties. Certainly, theory provides the conceptual basis from which experiments can be planned and analyzed. The calculations discussed here all... 

Even when the patterns are known to cluster, there remain difficult issues that must be addressed before a kernel-based approach can be used effectively. Two of the more fundamental conceptual issues are the number and size of clusters that should be used to characterize the pattern classes. These are issues for which there are no hard and fast answers. Despite the application of well-developed statistical methods, including squared-error indices and variance analysis, determining the number and size of clusters remains extremely formidable. 

A conceptually different approach to nanostructuring electrode surfaces by tipgenerated metal clusters is sketched in Figure 5.15h. This approach, which facilitates a so-called jump-to-contact between tip and substrate for generating metal clusters, has been developed by our group and will be described in more detail in Section 5.4.3. 

At shorter distances, particularly those characteristic of H-bonded and other charge-transfer complexes, the concepts of partial covalency, resonance, and chemical forces must be extended to intramolecular species. In such cases the distinction between, e.g., the covalent bond and the H-bond may become completely arbitrary. The concept of supramolecular clusters as fundamental chemical units presents challenges both to theory and to standard methods of structural characterization. Fortunately, the quantal theory of donor-acceptor interactions follows parallel lines for intramolecular and intermolecular cases, allowing seamless description of molecular and supramolecular bonding in a unified conceptual framework. In this sense, supramolecular aggregation under ambient thermal conditions should be considered a true chemical phenomenon. 

M. Quack and M. A. Suhm, Potential energy hypersurfaces for hydrogen bonded clusters (HF) , in Conceptual Perspectives in Quantum Chemistry, Conceptual Trends in Quantum Chemistry, Vol. Ill, J. L. Calais and E. S. Kryachko, ed., Kluwer, Dordrecht, 1997, pp. 415 463. 

---