Neural Networks

2012 ◽  
pp. 450-498 ◽  
Author(s):  
Siddhartha Bhattacharyya
Keyword(s):  
Neural Networks ◽  
Data Networks ◽  
Input Output ◽  
Training Set ◽  
Topology Preservation ◽  
Functional Characteristics ◽  
Output Data ◽  
Computing Paradigm ◽  
Mode Of Operation ◽  
Recent Trends

These networks generally operate in two different modes, viz., supervised and unsupervised modes. The supervised mode of operation requires a supervisor to train the network with a training set of data. Networks operating in unsupervised mode apply topology preservation techniques so as to learn inputs. Representative examples of networks following either of these two modes are presented with reference to their topologies, configurations, types of input-output data and functional characteristics. Recent trends in this computing paradigm are also reported with due regards to the application perspectives.

Flat Minima

1997 ◽  
Vol 9 (1) ◽  
pp. 1-42 ◽  
Author(s):  
Sepp Hochreiter ◽  
Jürgen Schmidhuber
Keyword(s):  
Neural Networks ◽  
Error Function ◽  
Low Complexity ◽  
Generalization Error ◽  
Input Output ◽  
Generalization Capability ◽  
Training Set ◽  
Weight Decay ◽  
Optimal Brain Surgeon ◽  
And Training

We present a new algorithm for finding low-complexity neural networks with high generalization capability. The algorithm searches for a “flat” minimum of the error function. A flat minimum is a large connected region in weight space where the error remains approximately constant. An MDL-based, Bayesian argument suggests that flat minima correspond to “simple” networks and low expected overfitting. The argument is based on a Gibbs algorithm variant and a novel way of splitting generalization error into underfitting and overfitting error. Unlike many previous approaches, ours does not require gaussian assumptions and does not depend on a “good” weight prior. Instead we have a prior over input output functions, thus taking into account net architecture and training set. Although our algorithm requires the computation of second-order derivatives, it has backpropagation's order of complexity. Automatically, it effectively prunes units, weights, and input lines. Various experiments with feedforward and recurrent nets are described. In an application to stock market prediction, flat minimum search outperforms conventional backprop, weight decay, and “optimal brain surgeon/optimal brain damage.”

scholarly journals Control on the Manifolds of Mappings with a View to the Deep Learning

2021 ◽  
Author(s):  
Andrei Agrachev ◽  
Andrey Sarychev
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Deep Learning ◽  
Control System ◽  
Continuous Time ◽  
Time Variable ◽  
Input Output ◽  
Training Set ◽  
Time Control ◽  
Interpolation Problems

AbstractDeep learning of the artificial neural networks (ANN) can be treated as a particular class of interpolation problems. The goal is to find a neural network whose input-output map approximates well the desired map on a finite or an infinite training set. Our idea consists of taking as an approximant the input-output map, which arises from a nonlinear continuous-time control system. In the limit such control system can be seen as a network with a continuum of layers, each one labelled by the time variable. The values of the controls at each instant of time are the parameters of the layer.

Non-linear time series analysis of combustion pressure data for neural network training with the concept of mutual information

2001 ◽  
Vol 215 (2) ◽  
pp. 299-304 ◽  
Author(s):  
F Heister ◽  
M Froehlich
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Mutual Information ◽  
Combustion Engine ◽  
Linear Time ◽  
Chamber Pressure ◽  
Neural Network Training ◽  
Input Output ◽  
Output Data ◽  
Network Training

In recent years, after a period of disillusion in the field of neural processing and adaptive algorithms, neural networks have been reconsidered for solving complex technical tasks. The problem of neural network training is the presentation of input/output data showing an appropriate information content which represent a given problem. The training of a neural structure will definitely lead to poor results if the relation between input and output signals shows no functional dependence but a pure stochastic behaviour. This paper is concerned with the identification of the most relevant input-output data pairs for neural networks, using the concept of mutual information. A general, quantitative method is demonstrated for identifying the most relevant points from the transient measured data of a combustion engine. In this context mutual information is employed for the problem of determining the 50 per cent energy conversion point solely from the combustion chamber pressure during one combustion cycle.

Extracting fuzzy rules for modeling of complex processes using neural networks

2013 ◽  
Vol 227 (12) ◽  
pp. 2861-2869 ◽  
Author(s):  
A Jamali ◽  
SJ Motevalli ◽  
N Nariman-zadeh
Keyword(s):  
Neural Networks ◽  
Fuzzy Rule ◽  
Structure Design ◽  
Fuzzy Rules ◽  
Group Method ◽  
Rule Base ◽  
Input Output ◽  
Output Data ◽  
Complex Processes ◽  
Linguistic Rules

Modeling of complex processes often leads to complex mathematical relationships between inputs and outputs, which do not reflect the influence of the independent variables on the output parameters. In this article, an innovative technique based on neural networks is presented to extract fuzzy linguistic rules for modeling some processes using some input–output data. In this way, genetic algorithm is used both for optimal structure design of those group method of data handling-type neural networks and for subsequent optimization of sub-bounds of fuzzy singleton antecedents to further optimize the obtained fuzzy rule base. Three different input–output data tables related to some complex problems of a nonlinear mathematical system, an explosive cutting process and the probability of failure estimation of a two mass-spring system are modeled by some fuzzy rules, using the technique discussed in this article.

scholarly journals Method of value chain mapping with international input–output data: application to the agricultural value chain in three Greater Mekong Subregion countries

2021 ◽  
Vol 10 (1) ◽  
Author(s):  
Ikuo Kuroiwa
Keyword(s):  
Value Chain ◽  
Value Added ◽  
Input Output ◽  
Output Data ◽  
Primary Input ◽  
Goods And Services ◽  
Unit Structure ◽  
Data Application ◽  
Final Output ◽  
Greater Mekong Subregion

AbstractExtending the technique of unit structure analysis, which was originally developed by Ozaki (J Econ 73(5):720–748, 1980), this study introduces a method of value chain mapping that uses international input–output data and reveals both the upstream and downstream transactions of goods and services, as well as primary input (value added) and final output (final demand) transactions, which emerge along the entire value chain. This method is then applied to the agricultural value chain of three Greater Mekong Subregion countries: Thailand, Vietnam, and Cambodia. The results show that the agricultural value chain has been increasingly internationalized, although there is still room to benefit from participating in global value chains, especially in a country such as Cambodia. Although there are some constraints regarding the methodology and data, the method proves useful in tracing the entire value chain.

Sign in / Sign up

Export Citation Format

Share Document