Learning to Design Analog Circuits to Meet Threshold Specifications

Overview

Code for Learning to Design Analog Circuits to Meet Threshold Specifications (Accepted by ICML 2023 as Poster)

• Preprint Coming Soon
• Website Comning Soon

Docker

To export running results to host machine:

1. Create two folders, one for out_plot and another for result_out.
2. Run the docker using command

docker run -v {absolute path to out_plot folder}:/Circuit-Synthesis/out_plot -v {absolute path to result_out folder}:/Circuit-Synthesis/result_out {docker image name} --path={Train config path}

Usage

Problem 1

Base dataset, DL, across data sizes with base test dataset, success rate as function of (two-sided) error margin:

python main.py --path=./config/config_template/problem1-compare-datasize-relative-Error-margin.yaml

Problem 2 Number 1 and Number 4

Compare datasets construction methods using deep learning, 10-fold Cross Validation as a function of error margin:

python main.py --path=./config/config_template/problem2-compare-dataset-DL-10fold-absolute-Error-margin.yaml

Problem 2 Number 2

Test success rate; Compare training methods (DL, lookup, RF, …) with “softargmax”, 10-fold cross validation as a function of error margin:

python main.py --path=./config/config_template/problem2-compare-method-Softargmax-10fold-absolute-Error-margin.yaml

Problem 2 Number 3

Test success rate; Compare data sizes with DL, “softargmax” as a function of error margin:

python main.py --path=./config/config_template/problem2-compare-datasize-softArgmax-DL-Absolute-Error-margin.yaml

Citation

(Coming soon)