where F is an estimate of some variable of interest, y, x = x1,...,xn and denotes the setof input variables or predictors, and ε is noise or an error term. The training of theneural network is analogous to parameter estimation in regression. As discussed inSection 1.2, neural networks can approximate any functional behavior, without theprerequisite a priori knowledge of the structure of the relationships that are described.As a result, numerous applications of predictive neural network models to environmentaland biological analyses have been reported in the literature. For example,neural networks have been incorporated into urban air quality studies for accurateprediction of average particulate matter (PM2.5 and PM10) concentrations in orderto assess the impact of such matter on the health and welfare of human populations(Perez and Reyes, 2006; Dong et al., 2009). They have also been widely incorporatedinto biological studies, for example, from the use of neural networks in predictingthe reversed-phase liquid chromatography retention times of peptides enzymaticallydigested from proteome-wide proteins (Petritis et al., 2003), to the diagnosis of heartdisease through neural network ensembles (Das et al., 2009).
đang được dịch, vui lòng đợi..