Sequential Monte Carlo Methods for Neural Networks
Many problems, arising in science and engineering, require the estimation of nonlinear, time-varying functions that map a set of input signals to a corresponding set of output signals. Some examples include: finding the relation between an input pressure signal and the movement of a pneumatic control valve; using past observations in a time series to predict future events; and using a group of biomedical signals to carry out diagnoses and prognoses. These problems can be reformulated in terms of a generic one of estimating the parameters of a suitable neural network on-line as the input-output data becomes available.
KeywordsExtended Kalman Filter Class Membership Importance Weight Proposal Distribution Sequential Monte Carlo
Unable to display preview. Download preview PDF.