Analog Hardware Neural Networks using Logarithmic Four-Quadrant Multiple Circuits

In the artificial intelligence, machine learning and neural network fields, many practical use models have been proposed. However, these models are based on a digital Von Neumann model's computer. There are a minority of studies in the field of analog learning neural networks. In our previous study, we used multiple circuits as connecting weights. Multiple circuits calculate the product of the two input values, input signals and connecting weights. However, the working range is very small because of the semiconductor characteristics. When constructing a network, one of the disadvantages is that the input and output range is limited. Furthermore, the circuit operation becomes unstable because of the characteristics of the multiple circuit. In this study, we proposed logarithmic four-quadrant multiple circuits as a connecting weight. From the experimental result, it is shown that this circuit operates stably over a wide range. The model works using pure analog electronic circuits. The learning time is quite short compare to a digital process computer. We improved the number of units and network layers. Moreover, we suggest the possibility of the realization of the hardware implementation of the deep learning model.