PDB-ANN

Table of Contents

• Overview
• Requirements
• Installation
• Usage
• Example
• License

Overview

The PDB-ANN project aims to implement one of the methodologies mentioned in the research paper - Optimal Search with Neural Networks: Challenges and Approaches.

Requirements

Before you begin, ensure you have met the following requirements:

• You have installed the latest version of LibTorch. Detailed instructions related to the installation of LibTorch can be found on the linked page.

• This project extensively utilizes the nlohmann/json library to facilitate the conversion of models into JSON format. You can download the latest version of this library via JSON for Modern C++.

• cxxopts is a lightweight C++ library used for parsing command-line arguments in this project. To get the latest version of cxxopts, visit the releases page.

Note: This project includes the nlohmann/json and cxxopts libraries, so there's no need to download them manually.

Installation

To build PDB-ANN, follow these steps:

Linux:

git clone https://github.com/mgh5225/PDB-ANN.git
cd PDB-ANN
bash build.sh

Please be aware that you'll need to update the absolute path to LibTorch in the CMakeLists.txt file.

Usage

Once the build process is complete, you can initiate the program using the following commands:

Pattern Database + ANN
Usage:
  main [OPTION...]

  -h, --help    Print usage
      --pdb     Create PDBs
      --create  Create random database based on created PDBs
  -t, --train   Train ANN
  -r, --run     Run ANN
  -f, --find    Find q*

 run options:
  -a, --all          Run on whole dataset
  -s, --state arg    State for ANN
  -p, --pattern arg  Pattern for ANN
  -d, --dim arg      Dimension for ANN
  -q arg             List of q for ANN

Example

the output of ./build/main -r --pattern 1,2,3,4 --state 8,7,3,0,4,5,1,6,2 --dim 3,3 -q 1e-2,1e-4,1 is as follows:

 5.5599e-12  5.7679e-04  9.9942e-01  4.1235e-09  8.6619e-09  2.9172e-10
[ CPUFloatType{1,6} ]
q       h
0.01    2
0.0001  1
1       2

License

This project uses the following license: MIT license