Analog Electronic Neural Networks: Analog Computing Combined with Digital Data Processing Revisited
Stefan Bosse, Björn Lüssem · 2024
Analog computing had its prime in the years from 1960 to 1970. With the rise of powerful digital computers, analog computing using transistor and Operational Amplififer (OPAMP) circuits vanished nearly completely but gained an increasing interest in recent years again. In this study, we will consider, in particular, the analog Artificial Neural Network (AANN) that can be considered as co-processors for digital systems. We will show that the training result of an ANN using digital algorithms can be transferred to analog transistor circuits. But this process is still a challenge and can fail. So, we discuss the limitations and possible solutions to generate and create an AANN from a modified digital model that takes limitations of electronic circuits into account.