We propose a two-stage training for the multilayer perceptron (MLP). The first stage is bottom-up, where we use a class separability measure to conduct hidden layer training and the least squared error criterion to train the output layer. The second stage is top-down, we use a criterion derived from classification error rate to further train the network weights. We demonstrate the effectiveness...

A binary classification problem is solved by acting on the combined evidence of several early vision modules. Each module gives an opinion as to the identity of an individual image element, and a consensus is reached by a trained Multi-Layer Perceptron (MLP).

A novel neural-network based technique is described for the remote condition-monitoring of an in-service gas-turbine flowmeter. The method uses a C language implementation of a modified multi-layer perceptron (MLP) neural networks, which enables detection of the accumulation of contaminating material on the rotor blades that could lead to changes in meter-factor and loss of calibration.

A method for automatic detection and localization of faces on digital images is proposed. The method is based on learning by example and multi-resolution analysis of digital images. Special emphasis is put on the management of the learning data, in order to improve the performances. Various experimental results, obtained by using a Multi-Layer Perceptron (MLP) as a classier, are provided.

Since artificial neural networks (ANNs) can approximate any function, they have been applied in many fields including hydrology. In hydrology, there are important issues such as flood estimation and predicting rainfall-runoff in a certain area. In this presentation, we briefly introduce a popular feed-forward neural network model, so called “multi-layer perceptron (MLP)”, and review its applica...

the main objective in classification of the nmr spectra of cancerous and healthy tissue , with high number of features is the prerequisites of the minimum number of samples. therefore the use of conventional classifier on this type of the data is not recommended. in the current work, different structures of the artificial neural networks (ann) were tried on classification of different cancerous...

Artificial Neural Network (ANN) learns from inputs and outputs. The values of the weights biases in ANN are updated according to Researchers have proposed algorithms train Multi-Layer Perceptron (MLP). However, classical techniques often face problems solving this optimization problem. They tend need large amounts computing time, memory. More importantly, they get stuck within local optimum pro...

A general problem in model selection is to obtain the right parameters that make a model t observed data. If the model selected is a Multilayer Perceptron (MLP) trained with Backpropagation (BP), it is necessary to nd appropriate initial weights and learning parameters. This paper proposes a method that combines Simulated Annealing (SimAnn) and BP to train MLPs with a single hidden layer, terme...

In the area of information processing one fundamental issue is how to measure the statistical relationship between two variables based only on their samples. In a previous paper, the idea of Information Potential which was formulated from the so called Quadratic Mutual Information was introduced, and successfully applied to problems such as Blind Source Separation and Pose Estimation of SAR (Sy...