Hybrid ATPG (Automatic Test Pattern Generation) algorithm

The project \"NN-for-ATPG\" proposes a hybrid approach for Automatic Test Pattern Generation (ATPG) by integrating machine learning with the FAN algorithm. It includes implementations of two approaches, NN-Hyb and NN-All. It features two implementations, NN-Hyb and NN-All, which apply neural network models at selective and all circuit levels, respectively. Project files support the paper by providing the source code, circuit files, and scripts needed to reproduce key results, including comparisons for \"all fault cases,\" \"hard-to-detect faults,\" and runtime comparisons with and without acceleration. Minimal requirements include a C++ compiler and shell script execution capabilities., , , ## File Structure This repository includes the following directories and files (full file for download under Related works 'Software'): * `README.md`: instructions to compile and generate results in the paper * `NN-Hyb`: directory containing source code of our proposed approach * `NN-All`: directory containing soruce code of alternative approach compared in the paper * `circuits`: circuit files used in our experiments (also found on Dryad) * `hard_faults`: hard-to-detect faults per circuit used in results of Table II in the paper * `script`: scripts to generate results in Tables I, II, III in the paper ## Installation In the paper we have compared the following 3 approaches: (1) Fan: base approach for ATPG used for comparison (2) NN-Hyb: our approach based on applying our neural network (NN) model during backtrace at select levels of the circuit during ATPG (3) NN-All: another alternative approach using our NN model, but at all levels of the circuit To get an executable for (1), p...项目“NN-for-ATPG”提出了一种将机器学习与FAN算法相结合的自动测试向量生成（Automatic Test Pattern Generation, ATPG）混合方法，包含NN-Hyb与NN-All两种实现方案。这两种方案分别在电路的选择性层级与全电路层级应用神经网络模型。项目文件提供了复现关键结果所需的源代码、电路文件与脚本，可支撑对应论文的研究成果，涵盖“全故障案例”“难检测故障”的对比实验，以及带加速与不带加速的运行时对比。项目的最低运行要求为C++编译器与Shell脚本执行环境。 ## 文件结构 本仓库包含以下目录与文件（完整下载文件见相关作品「Software」板块）： * `README.md`：编译与复现论文中结果的操作指南 * `NN-Hyb`：存放本项目提出方法的源代码目录 * `NN-All`：存放论文中对比的替代方案源代码的目录 * `circuits`：本实验所用的电路文件（亦可在Dryad平台获取） * `hard_faults`：论文表II结果中所用的各电路难检测故障集合 * `script`：用于生成论文表I、表II、表III中结果的脚本 ## 安装 本论文中对比了以下三种方法： (1) FAN：用于对比的基础ATPG方法 (2) NN-Hyb：本项目提出的方法，在ATPG过程中针对电路的选择性层级在回溯阶段应用神经网络模型 (3) NN-All：另一款对比方案，在电路的全层级应用神经网络模型 若要获取(1)的可执行文件，请参阅……