MW01b-22 Hardware Implementation of Neuromorphic Circuits

Subscribe
Notify of
4 Comments
Inline Feedbacks
View all comments
ZHENG, Zheyang
January 18, 2023 2:34 pm

Pasting your paper in a poster is not a usual practice. Please highlight:

  1. What kind of hardware do you use?
  2. What are the expected functions of your neuromorphic circuits?
  3. What are the algorithms and how does the hardware implementation benefit?
XU, Zimo
January 18, 2023 2:59 pm
Reply to  ZHENG, Zheyang
  1. What kind of hardware do you use?

Operational amplifiers OPA1622; Decoder CD4514BM96; Voltage amplifier LM324M; Capacitors; Data Selector CD4051BPWR; A series of adjustable resistors; Dual-gate thin-film transistor; STM32F407ZGT6.

ZHANG, Zhejun
January 18, 2023 3:19 pm
Reply to  ZHENG, Zheyang

What are the expected functions of your neuromorphic circuits?
Ans: Actually, the neuromorphic circuit is a very new topic. For the current stage, this neuromorphic can execute some logic functions, such as OR, AND, NULL, IDENTITY function, and so on. Unfortunately, the logic functions, which include NOT, are feasible to execute due to the difficulty in the training part. But we will keep going through more actual functions in the spring semester. Such as the Convolution function, we expected that the circuits could find the best weight voltage signal by utilizing the expected example output and gradient function. And then, It could execute the corresponding type of filter by using a hardware neuromorphic circuit only.

XU, Zimo
January 18, 2023 3:29 pm
Reply to  ZHENG, Zheyang
  1. What are the algorithms and how does the hardware implementation benefit?

We are using the gradient descent algorithm. The algorithm will lead to a significant decrease of calculation time and find the minimum cost more effectively.