Pattern Recognition

o Neural filters for radiation pattern analysis

    Developed a feed forward neural network to  distinguish between true
    and false radiation signals and to estimate particle energies in the
    presence  of  a  very  high  level  of  noise  for  certain  nuclear
    reactions.  The  network  consists  of  an  input  layer, two hidden
    layers,  and  an  output  layer.   The  first  layer connections are
    developed by using the  organization  principles  found in low-level
    mammalian visual system:  maximizing  information transmission.  The
    connections of the second and third layers are learned by back error
    propagation on the  first  layer  output.   The  network learns from
    training data to do pattern recognition in raw  input particle data.
    It is shown that the network performs  the  recognition  task with a
    high degree of accuracy, higher than trained human eyes.

  ^{Dong D W and Gyulassy M 1992}
    {Neural network approach to process jet fragmentation information}
    {Proc International Joint Conference on Neural Networks, Baltimore}
    { Vol~3 pp~191-196}

  ^{Dong D W and Gyulassy M 1993}
    {Neural filters for jet analysis}
    {(LBL-31560) Physical Review E}{ Vol~47(4) pp~2913-2922}

  ^{Dong D W and Chan Y D 1993}
    {Three layer network for identifying Cerenkov radiation patterns}
    {Proc World Congress on Neural Networks, Portland}{ Vol~1 pp~312-315}

  ^{Dong D W and Chan Y D 1993}
    {Neural network for recognizing Cerenkov radiation patterns}
    {LBL-33634}{}

o Neural networks for engine fault diagnostics

    A dynamic neural network  is  developed  to  detect  soft failures of
    sensors and actuators in automobile engines.   The network, currently
    implemented off-line in software, can process multi-dimensional input
    data in real time.  The network  is  trained  to  predict  one of the
    variables using others.  It learns  to  use  redundant information in
    the variables such as higher order statistics and temporal relations.
    The difference between the prediction and the measurement  is used to
    distinguish a normal engine from a faulty one.  Using the network, we
    are  able  to  detect  errors  in  the  manifold  air pressure sensor
    (Vs) and the exhaust gas recirculation valve (Va) with  a high degree
    of accuracy.

  ^{Dong D W, Hopfiled J J, and Unnikrishnan K P 1997}
    {Neural networks for engine fault diagnostics}
    {Neural Networks for Signal Processing}
    { Vol~VII pp~636-644}

( Papers' Index of Dawei Dong )


Send comments to Dawei Dong: dawei@ccs.fau.edu