Evaluation of Parameter Settings for Training Neural Networks Using Backpropagation Algorithms

Artificial neural networks (ANN) are widely used for classification, and the training algorithm commonly used is the backpropagation (BP) algorithm. The major bottleneck faced in the backpropagation neural network training is in fixing the appropriate values for network parameters. The network parameters are initial weights, biases, activation function, number of hidden layers and the number of neurons per hidden layer, number of training epochs, learning rate, minimum error, and momentum term for the classification task. The objective of this work is to investigate the performance of 12 different BP algorithms with the impact of variations in network parameter values for the neural network training. The algorithms were evaluated with different training and testing samples taken from the three benchmark clinical datasets, namely, Pima Indian Diabetes (PID), Hepatitis, and Wisconsin Breast Cancer (WBC) dataset obtained from the University of California Irvine (UCI) machine learning repository.

Random fiber network loaded by a point force

This article presents the displacement field produced by a point force acting on an athermal random fiber network (the Green function for the network). The problem is defined within the limits of linear elasticity and the field is obtained numerically for nonaffine networks characterized by various parameter sets. The classical Green function solution applies at distances from the point force larger than a threshold which is independent of the network parameters in the range studied. At smaller distances, the nonlocal nature of fiber interactions modifies the solution.

A Novel Theoretical Probabilistic Model for Opportunistic Routing with Applications in Energy Consumption for WSNs

This paper proposes a new theoretical stochastic model based on an abstraction of the opportunistic model for opportunistic networks. The model is capable of systematically computing the network parameters, such as the number of possible routes, the probability of successful transmission, the expected number of broadcast transmissions, and the expected number of receptions. The usual theoretical stochastic model explored in the methodologies available in the literature is based on Markov chains, and the main novelty of this paper is the employment of a percolation stochastic model, whose main benefit is to obtain the network parameters directly. Additionally, the proposed approach is capable to deal with values of probability specified by bounded intervals or by a density function. The model is validated via Monte Carlo simulations, and a computational toolbox (R-packet) is provided to make the reproduction of the results presented in the paper easier. The technique is illustrated through a numerical example where the proposed model is applied to compute the energy consumption when transmitting a packet via an opportunistic network.

Empirical Analysis of Financial Depth and Width Based on Convolutional Neural Network

There are great differences in financial and economic development in different regions. In different time series and different regions, the effects of financial depth and width on economic development are also different. This paper selects neural network to establish the economic benefit model of financial depth and breadth, which can deeply explore the relationship between financial data and economic data. In order to determine the optimal convolutional neural network parameters, the optimal convolutional neural network parameters are determined through comparative simulation analysis. The convolutional neural network model based on the optimal parameters is applied to the empirical analysis of the effect of financial and economic development in X region. In order to obtain the optimal convolutional neural network parameters, different convolution layers, convolution core size, and convolution core number are compared and simulated. The convolutional neural network model with optimal parameters is used to simulate the financial and economic data of X region. The simulation results show that the density of financial personnel has a certain impact on economic development, so it is necessary to improve the comprehensive quality of financial personnel and promote regional economic development. Therefore, this paper seeks an effective method to study the effect of financial breadth and depth on economic development which can provide a feasible idea for the in-depth research method of financial and economic development.

Compensatory variability in network parameters enhances memory performance in the Drosophila mushroom body

Neural circuits use homeostatic compensation to achieve consistent behavior despite variability in underlying intrinsic and network parameters. However, it remains unclear how compensation regulates variability across a population of the same type of neurons within an individual and what computational benefits might result from such compensation. We address these questions in the Drosophila mushroom body, the fly’s olfactory memory center. In a computational model, we show that under sparse coding conditions, memory performance is degraded when the mushroom body’s principal neurons, Kenyon cells (KCs), vary realistically in key parameters governing their excitability. However, memory performance is rescued while maintaining realistic variability if parameters compensate for each other to equalize KC average activity. Such compensation can be achieved through both activity-dependent and activity-independent mechanisms. Finally, we show that correlations predicted by our model’s compensatory mechanisms appear in the Drosophila hemibrain connectome. These findings reveal compensatory variability in the mushroom body and describe its computational benefits for associative memory.

Selection of features system and network parameters for hyperspectral images classification using convolutional neural networks

A classification system for hyperspectral images using convolutional neural networks is described. A specific network was selected and analyzed. The network parameters, ensured the maximum classification accuracy: dimension of the input layer, number of the layers, size of the fragments into which the classified image is divided, number of learning epochs, are experimentally determined. High percentages of correct classification were obtained with a large-format hyperspectral image, and some of the classes into which the image is divided are very close to each other and, accordingly, are difficult to distinguish by hyperspectra.