Feasibility of synchronous pulse-density modulation arithmetic in integrated circuit implementations of artificial neural networks

The major goal is to find an effective method to implement artificial neural network structures on VLSI circuits. Pulse-density modulation (PDM) arithmetic offers a new interesting method to implement artificial neural network structures on silicon. The advantage is simple arithmetic structures leading to small area on silicon and thus larger networks on a single circuit. The pulse-density modulation technique in a restricted sense mimics the biological idea of neuron action using both analog and digital structures. The synchronous PDM-technique, PDM arithmetic and structures, and VLSI implementations are discussed. The data format causes some new restrictions for the overall architecture of the circuits