102,003 research outputs found
Clustering in Remote Sensing Using an Unsupervised Neural Network
No full textAn unsupervised neural network based on a new neural unit is applied to the problem of clustering a data set of pixels drawn through remote sensing of a limited portion of the terrestrial surface. The aim of the partition is to split the sensed area into sets having roughly the same type of ground cover. After the partitioning a comparison with the ground truth has shown that the unsupervised net has been able to split accurately the given data set in subsets similar to the classes really observable with a fast convergence and a high resolution
The topological foundations of electrical networks — complete models
No full textThe aim of this paper is to identify the topological entities underlying the construction of a complete (not reduced) network model. Consider an active network as a pair of physically different structures, i.e. a ‘dead network’ and an ‘excitation’. First, we define the general structure of the excitation which can be applied to a dead network; then the graphs Ga and G of the active and dead networks, respectively, are introduced. It is shown that, if a complete network model is to be constructed, the topological relations among the voltages and those among the currents of the network elements must be written choosing maximal independent sets of loops and cut‐sets of Ga. It is successively pointed out that the graph G can be obtained from Ga by means of a reduction operation, and the topological entities of G corresponding to the loops and to the cut‐sets of Ga are singled out. It has been found that a complete model of an electric network implies the use of four topological entities defined on the set of G‐edges, i.e. the closed path, the open path, the cut‐set and the pseudo‐cut, together with two topological entities defined on a maximal set of independent vertex couples of G, i.e. the ‘junction pair’ extreme of an open path and the ‘family of independent node couples’ split by a cut‐set. Furthermore the different reference frames and the relative topological matrix transformations which allow a complete model of the network to be built up are singled out and discussed
Heteroassociative memories via globally asymptotically stable discrete time network cellular
No full textOrganizing rules production in qualitative modeling of electric circuits
No full textThe prototype of a knowledge base (KB) for qualitative analysis and a method to create the set of rules necessary to examine an electrical circuit, starting with its topological description, are presented. Rules are grouped in functional blocks and two aggregation criteria allow one to automatically extract and manage suitable blocks. In this way a detailed proper KB is created for qualitatively analyzing any considered circuit when submitted to a backward chaining inference engine
A Multiple Neural Network System to Classify Solder Joints on Integrated Circuits
No full textThe following paper introduces a diagnostic
process to detect solder joint defects on Printed Circuit Boards
assembled in Surface Mounting Technology. The diagnosis is
accomplished by a Neural Network System which processes the
images of the solder joints of the integrated circuits mounted on
the board. The board images are acquired and then preprocessed
to extract the regions of interest for the diagnosis
which are the solder joints of the integrated circuits. Five
different levels of solder quality in respect to the amount of
solder paste have been defined. Two feature vectors have been
extracted from each region of interest, the “geometric” feature
vector and the “wavelet” feature vector. Both vectors feed the
neural network system constituted by two Multi Layer
Perceptron neural networks and a Linear Vector Quantization
network for the classification. The experimental results are
devoted to comparing the performances of a Multi Layer
Perceptron network, of a Linear Vector Quantization network,
and of the overall neural network system, considering both
geometric and wavelet features. The results prove that the
overall classifier is the best compromise in terms of recognition
rate and time required for the diagnosis in respect to the single
classifiers
- …
