In an on-line learning environment where optimal recognition performance over the newly encountered patterns is required , a robust incremental learning procedure is necessary to re-conngure the entire neural network without aaecting the stored information. In this paper, an heuristic pattern correction scheme based upon an hierarchical data partitioning principle is proposed for digit word rec...

We present a hybrid visualization technique for compound graphs (i.e., networks with a hierarchical clustering defined on the nodes) that combines the use of adjacency matrices, node-link and arc diagrams to show the graph, and also combines the use of nested inclusion and icicle diagrams to show the hierarchical clustering. The graph visualized with our technique may have edges that are weight...

altan derivatives of polycyclic conjugated hydrocarbons were recently introduced and studied in theoretical organic chemistry. we now provide a generalization of the altan concept, applicable to any graph. several earlier noticed topological properties of altan derivatives of polycyclic conjugated hydrocarbons are shown to be the properties of all altan derivatives of all graphs. among these ar...

let g be a (p,q) graph. an injective map f : e(g) → {±1,±2,...,±q} is said to be an edge pair sum labeling if the induced vertex function f*: v (g) → z - {0} defined by f*(v) = σp∈ev f (e) is one-one where ev denotes the set of edges in g that are incident with a vertex v and f*(v (g)) is either of the form {±k1,±k2,...,±kp/2} or {±k1,±k2,...,±k(p-1)/2} u {±k(p+1)/2} according a...

Andrew R. Barron Department of Statistics Yale University P.O. Box 208290 New Haven, CT 06520 Andrew. Barron@yale. edu Gaussian mixtures (or so-called radial basis function networks) for density estimation provide a natural counterpart to sigmoidal neural networks for function fitting and approximation. In both cases, it is possible to give simple expressions for the iterative improvement of pe...

We consider the nonlinear Schrödinger equation iut = −∆u−|u|p−1u in dimension N ≥ 3 in the L super critical range N+3 N−1 ≤ p < N+2 N−2 . The corresponding scaling invariant space is Ḣc with 1 2 ≤ sc < 1 and this covers the physically relevant case N = 3, p = 3. The existence of finite time blow up solutions is known. Let u(t) ∈ Ḣc ∩ Ḣ be a radially symmetric blow up solution which blows up at ...

In this paper a handwritten recognition algorithm based on structural characteristics, histograms and profiles, is presented. The well-known horizontal and vertical histograms are used, in combination with the newly introduced radial histogram, out-in radial and inout radial profiles for representing 32x32 matrices of characters, as 280-dimension vectors. The recognition process has been suppor...

Human vision is often guided by instinctual commonsense such as proportions and contours. In this paper, we explore how to use the proportion as the key knowledge for designing a privacy algorithm that detects human private parts in a 3D scan dataset. The Analogia Graph is introduced to study the proportion of structures. It is a graph-based representation of the proportion knowledge. The intri...

Visualizing dynamic graphs is challenging because changing node and edge attributes as well as topological alterations need to be encoded in the visual representation. However, existing approaches such as animation, juxtaposition, and superimposition do not scale well. In this poster we propose a novel layering approach for visualizing dynamic graphs where the graph for each point in time is a ...

We construct center-stable and center-unstable manifolds, as well as stable and unstable manifolds, for the nonlinear Klein–Gordon equation with a focusing energy subcritical nonlinearity, associated with a family of solitary waves which is generated from any radial stationary solution by the action of all Lorentz transforms and spatial translations. The construction is based on the graph trans...