Rotation, size and shape recognition by a spreading neural network
Keisuke Nakamura, S. Miyamoto, T. Yoshikawa · 2003
Proposes a rotation and size spreading associative neural network that recognises the size, orientation and shape of an object. This neural net uses spatial spreading by means of a double spreading layer, generalised inverse learning and population vector method for recognition of the objects. This neural net can simultaneously recognise the size of an object (irrespective of its orientation and shape), its orientation (irrespective of its size and shape) and its shape (irrespective of its size and orientation). The neural net spreads the information about the orientation and size of the object by double-spreading weights which have similar tuning characteristics to the axis-orientation neurons and size-discrimination neurons in the parietal cortex.