Bicomplex Neural Networks with Hypergeometric Activation Functions

Abstract Bicomplex convolutional neural networks (BCCNN) are a natural extension of the quaternion for bicomplex case. As it happens with quaternionic case, BCCNN has capability learning and modelling external dependencies that exist between neighbour features an input vector internal latent within feature. This property arises from fact that, under certain circumstances, is possible to deal number in component-wise way. In this paper, we present BCCNN, apply classification task involving colourized version well-known dataset MNIST. Besides novelty considering numbers, our CNN considers activation function Bessel-type function. see, results better compared one where classical ReLU considered.