QuasiNet: a neural network with trainable product layers

arXiv:2401.06137v2 Announce Type: replace-cross
Abstract: Classical neural networks achieve only limited convergence in hard problems such as XOR or parity when the number of hidden neurons is small. With the motivation to improve the success rate of neural networks in these problems, we propose a new neural network model inspired by existing neural network models with so called product neurons and a learning rule derived from classical error backpropagation, which elegantly solves the problem of mutually exclusive situations. Unlike existing …

abstract arxiv convergence cs.ai cs.lg cs.ne hidden motivation network networks neural network neural networks neurons product rate small success type xor