Design of Analog Neuron Circuit with ReLU Activation Function
Jiayi Zhang, Xusen Zhao, Weijie Zhu · 2025
This paper presents an innovative analog neuron circuit design, subtly implementing ReLU activation function. This analogy neuron mainly consists of two parts. The first part is about linear processing, which responses for weighted summation of input signals. In our design, with the use of hardware circuit, it is able to assign desired weight to signals from different inputs. The other part is activation function circuit. In this part, we designed four different circuits. Proteus simulation results indicate that all four circuits can implement ReLU function well. With precision rectifier circuit selected, the output voltage has a strong ReLU relationship with the input, while the other circuits have their advantages and disadvantages. Additionally, the diode clamp circuit realizes ELU function approximately, which provides a new idea for the design of analog neurons.