Node-building parameter

In more complex problems, where the classes are more similar, the groups might not be distinct and similar patterns would be joined together. The links in this case would be of medium length. This is a desirable result because it is often interesting to find classes that are similar within the data-set. If dissimilar classes are expected, adjustment of the node-building parameter and the cluster threshold would be the first priority. 

Parameters related to the host framework nodes, connections, coordination polyhedra, secondary building units, topology, composition, bonds and charges. 

For m input variables the pseudo-dimension for prediction by a multilayer perceptron neural network requires that at least m+1 independent samples are available per node for building a model (Sontag, 1998 Schmitt, 2001). It therefore appears that a larger set of data points is required to fit nonlinear models, such as neural networks that generally have a large number of parameters (weights) to fit. 

Having prepared an evolutionary optimization strategy, the application to earthquake resistance follows without many problems. We model the building by the 8 dof/node elements, define a goal or fitness function - a function of the building s response to the earthquake excitation - and try to minimise the goal wrt. the compensators parameters. 

We remark that Theorem 12.1 remains valid also in the case where the separatrix enters an edge of the node region, i.e. F C. However, a complete bifurcation analysis in this case requires an additional governing parameter. It is introduced in the following way. Let us build a cross-section Sq to the on-edge homoclinic loop F, i.e. we define So = y = d, a < d/2, as depicted in Fig. 12.1.4. At the bifurcation point, the separatrix F intersects Sq at some... 

---