K. Schadler and F. Wysotzki. Application of a neural net in classification and knowledge discovery. ESANN 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....characterizing relations between redistribution problems. To capture mappings between different concepts (as heat water; Falkenhainer et al. 1989) or relations (as :IH 3 ; Anderson Thompson, 1989) the definition can be relaxed to bijective mappings between (similar) node and or arc labels (Schadler Wysotzki, 1998). In the following, we investigate a small subset of the possible variations of types and degrees of partial source target isomorphisms. 3 Experiment In a first experiment not reported in detail here we explored the suitability of our domain for studying transfer of non isomorphic ....

Schadler, K., & Wysotzki, F. (1998). Application of a neural net in classification and knowledge discovery. In Proc. of the 3rd Int. Workshop of Neural Networks in Applications (NN'98) (p. 219-226). Magdeburg.

....we allow two different label sets for G and G 0 . That is, we define weak isomorphisms. For partial isomorphisms between problems we might be interested in the size of the isomorphical part between structures. This is captured by measures of graph distance as for example defined by (Schadler Wysotzki, 1998). We will use a simplified version of Schadler s graph distance (see def. 2) Definition 2 (Distance between graphs) The distance between two graphs G and H can be defined as d(G; H) 1 Gamma VGH NGH max(VG ; VH ) max(NG ; NH ) with VGH as number of common arcs, NGH as number of common ....

Schadler, K., & Wysotzki, F. (1998). Application of a neural net in classification and knowledge discovery. In Neural Networks in Applications NN'98, Proceedings of the Third International Workshop (p. 219-226). Magdeburg.

....characterizing relations between redistribution problems. To capture mappings between different concepts (as heat water; Falkenhainer et al. 1989) or relations (as = Gamma; Anderson Thompson, 1989) the definition can be relaxed to bijective mappings between (similar) node and or arc labels (Schadler Wysotzki, 1998). In the following, we investigate a small subset of the possible variations of types and degrees of partial source target isomorphisms. 3 Experiment In a first experiment not reported in detail here we explored the suitability of our domain for studying transfer of non isomorphic ....

Schadler, K., & Wysotzki, F. (1998). Application of a neural net in classification and knowledge discovery. In Proc. of the 3rd Int. Workshop of Neural Networks in Applications (NN'98) (p. 219-226). Magdeburg.

.... (graph matchings) The problem of computing the largest (induced) subgraph U , i.e. maximizing the number jV U j of vertices, which can be embedded in two other graphs G and H is often mapped to the problem of computing the maximum clique in a vertex compatibility graph (VCG) see e.g. [7]. The maximum clique problem is known to be NP hard. In the case of a VCG the size of the problem depends highly on the definition of vertex compatibility. Compatibility of equally labeled vertices leads to a smaller VCG compared to the case that all vertices can be mapped onto each other. We ....

K. Schadler and F. Wysotzki. Application of a neural net in classification and knowledge discovery. In Proc. ESANN'98, 1998.

....the proximities between the input and the particular prototype graph. The classifier uses the proximity data to assign the input graph to one of the finite number of classes represented by the corresponding prototypes. The most common and classical approach uses a maximum selector as a classifier ([14], 15] The maximum selector classifier determines the maximum value from a set of numerical input data by directly comparing the values and finally selects the class corresponding to the largest value. The task of retrieving the most similar prototype is a basic problem not only in ....

.... of the energy function ( 16] For our purposes we choose as a representative model a variant of the algorithm from ( 18] since it is simple and it has not only been applied to classical optimization problems, but also to practical classification and knowledge discovery tasks in chemistry ([14], 15] Thus, we make the following design decisions: The transfer function f is the piecewise linear limiter function f(x) 8 : 1 : if x 1 x : if x 2 ] 0; 1 [ 0 : if x 0 (2) with an upper and lower saturation point. The selfinhibition d is set to zero, such that each unit i has a ....

K. Schadler and F. Wysotzki. Application of a neural net in classification and knowledge discovery. In M. Verleysen, editor, Proc. ESANN'98, pages 117--122. D-Facto, Brussels, 1998.

....This model agrees with other models of competitive processes and the formation of structures in the human brain (c.f. 62] and allows to model the strength of relaxable constraints locally as the weights of connections. In our applications, a neural net approach for graph matching developed in [63, 64, 60, 65] has been used. A pair of graphs can be transformed into a neural net whose steady states describe both a good mapping between the graphs and, by using the energy of a state corresponding to a local energy minimum, an upper bound for the distance between the graphs. The structure of the net is ....

Kristina Schadler and Fritz Wysotzki. Application of a neural net in classification and knowledge discovery. In Michael Verleysen, editor, Proc. ESANN'98, pages 117--122. D-Facto, 1998.

No context found.

K. Schadler and F. Wysotzki. Application of a neural net in classification and knowledge discovery. ESANN 1998.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC