ANALOG HARDWARE MODEL FOR MORPHOLOGICAL NEURAL NETWORKS

摘要

This paper presents a discrete analog hardware model for the morphological neural network. Morphological Neural Networks (MNN) are a new type of neural networks described by Ritter, et.al. These types of neural networks replace the classical operations of multiplication and addition by addition and maximum or minimum operations. The maximum and minimum operations allow to perform a nonlinear operation before the application of the activation or transfer function. MNN utilize algebraic lattice operations structure known as semi-ring (R_(+-∞) , ∨,∧ , +, + '), different from traditional neural networks that are based on the algebraic structure known as ring (R,+,x). The operations ∧ and ∨ denote binary operations of minimum and maximum, respectively. The hardware model is implemented using diodes, resistors, and 741 op-amps. The model has been tested using Pspice and implemented using analog components to evaluate its performance. The results show that the model is easy to implement and accurate results are obtained similar to the theoretical computational model. Physical properties of the components are evaluated to study the model ability to implement the computational neuron. The hardware implementation will be very useful in applications where a computer is not required or too expensive to justify its use. This type of design could be later implemented using integrated circuit technology.