UNIVERSAL APPROXIMATION WITH NON-SIGMOID HIDDEN LAYER ACTIVATION FUNCTIONS BY USING ARTIFICIAL NEURAL NETWORK MODELING
Journal Title: Indian Journal of Computer Science and Engineering - Year 2014, Vol 5, Issue 5
Abstract
Neural networks are modeled on the way the human brain. They are capable of learning and can automatically recognize by skillfully training and design complex relationships and hidden dependencies based on historical example patterns and use this information for forecasting. The main difference, and at the same time is biggest advantage of the model of neural networks over statistical techniques seen that the forecaster the exact functional structure between input and Output variables need not be specified, but this by the system with certain Learning algorithms is "learned" using a kind of threshold logic. Goal of the learning procedure is to define the training phase while those parameters of the network, with Help the network has one of those adequate for the problem behavior. Mathematically, the training phase is an iterative, converging towards a minimum error value process. They identify the processors of the network, minimize the "total error". The currently the most popular and most widely for business applications algorithm is the backpropagation algorithm. This paper opens the black box of Backpropagation networks and makes the optimization process in the network over time and locally comprehensible.
Authors and Affiliations
R. Murugadoss , Dr. M. Ramakrishnan
Keywords
Related Articles
- EP ID EP142353
- DOI -
- Views 117
- Downloads 0
Comparative study on classification power of the attributes and reducts
In rough set theory, reduct are those attributes which are important and are able to represent the whole range of attributes. They show up the important features of database. In the present study of educational data mini...
A REMOTE LOGIN AUTHENTICATION SCHEME WITH SMART CARDS BASED ON UNIT SPHERE
On the basis of the fact that points on a unit sphere are easy to compute but hard to derive, this paper presents a remote login authentication scheme with smart card based on unit sphere. The proposed scheme is simple,...
Visual Query in Temporal Database with Multi Dimensions
A wide range of database applications manage time-varying data. In contrast, existing database technology provides little support for managing such data. The research area of temporal databases aims to change this state...
An Analytical Study of Different Image Inpainting Techniques
Inpainting is the technique of filling in holes in an image to preserve its overall continuity. Applications of this technique include the restoration of old photographs and damaged film; removal of superimposed text lik...
RANK LEVEL FUSION USING FINGERPRINT AND IRIS BIOMETRICS
Authentication of users is an essential and difficult to achieve in all systems. Shared secrets like Personal Identification Numbers (PIN) or Passwords and key devices such as Smart cards are not presently sufficient in...