We propose a characteristic structure number interrelating the weight vectors of a feed-forward neural network. It allows the monitoring of the learning process of feedforward neural networks and the identiication of characteristic points/phases during the learning process. Some properties are given and results of applications to diierent networks are shown.

Applications of multi-layer feed-forward artificial neural networks (ANN) to spectroscopy are reviewed. Network architecture and training algorithms are discussed. Backpropagation, the most commonly used training algorithm, is analyzed in greater detail. The following types of applications are considered: data reduction by means of neural networks, pattern recognition, multivariate regression, ...

In this paper, we aim at developing scalable neural network-type learning systems. Motivated by the idea of “constructive neural networks” in approximation theory, we focus on “constructing” rather than “training” feed-forward neural networks (FNNs) for learning, and propose a novel FNNs learning system called the constructive feed-forward neural network (CFN). Theoretically, we prove that the ...

8 The training of some types of neural networks leads to separable non-linear least squares problems. These problems may be 9 ill-conditioned and require special techniques. A robust algorithm based on the Variable Projections method of Golub and Pereyra 10 is designed for a class of feed-forward neural networks and tested on benchmark examples and real data.

The paper presents an application of feed-forward neural networks: the simulation of combinational logical circuits (CLC). In the first part a presentation of main problems of CLC design is displayed, followed, in the second part by an example of simulating such circuits by artificial neural networks.

A synthesis of sliding mode control for a robot arm based on a robot model identified by neural networks is presented. The proposed structural feed-forward neural networks estimate the elements of the Lagrange-Euler mathematical model of robot, and they can be directly used for a synthesis of a model-based control system.

This paper evaluates the performance of boosted decision trees for tagging b-jets. It is shown, using a Monte Carlo simulation of WH → lνqq̄ events that boosted decision trees outperform feed-forward neural networks. The results show that for a b-tagging efficiency of 90% the b-jet purity given by boosted decision trees is almost 20% higher than that given by neural networks.

2 Description of Progress and Implementation 2.1 Implementation Our implementation is based on David Ackley and Michael Littman’s ERL. Each agent in the world has two neural networks: an action network and an evaluation network. The neural networks are implemented as feed-forward neural networks (see figure below) which have a set of input neurons, a number of hidden layers each containing hidd...

We consider image transformation problems, where an input image is transformed into an output image. Recent methods for such problems typically train feed-forward convolutional neural networks using a per-pixel loss between the output and ground-truth images. Parallel work has shown that high-quality images can be generated by defining and optimizing perceptual loss functions based on high-leve...

The nervous system encodes continuous information from the environment in the form of discrete spikes, and then decodes these to produce smooth motor actions. Understanding how spikes integrate, represent, and process information to produce behavior is one of the greatest challenges in neuroscience. Information theory has the potential to help us address this challenge. Informational analyses o...