A High Performance Neurocomputing Algorithm for Prediction Tasks in Wireless Sensor Networks

Equidistant Piecewise function Approximation for neurocomputing based environmental monitoring in wireless sensor networks

“…In order to improve the algorithm, this crucial algebraic term is simplified. As shown in [2], substituting the sigmoid function by linear regression equivalents is speeding up the algorithm runtime and barely causes loss of accuracy. In more detail, the activation function is simplified in this case to…”

“…An additional benefit attached to these changes is the use of the fixed point numeric format which is possible, as only common operands are required in this approximation case. As described in [2], the use of linear grade polynomials delivers best results in case of environmental monitoring for container based perishable food transportation. Thus, this approximation is used as reference for the EPA approach whose main principle is described in detail in the following.…”

“…In order to increase the battery lifetime, prediction algorithms are used that enable function course prediction of environmental data [5]. In general, there are several different mathematical approaches for the prediction task that vary mainly in runtime, accuracy and sampling flexibility [2].…”

“…To minimize this disadvantage, related work focused on simplification of mathematical terms by approximation means [2]. In more detail, linear regression is used to replace crucial algebraic equations.…”

“…Its main advantage is high accuracy prediction of environmental parameters, such as temperature or humidity [2]. Although predictors reduce in general the transceiver energy, their corresponding algorithm requires high calculation effort that may nullify this benefit.…”