See the wiki for the quick start guide.
HyperMapper is a multi-objective black-box optimization tool based on Bayesian Optimization.
HyperMapper was succesfully applied to real-world problems involving design search spaces with trillions of possible design choices. In particular it was applied to:
- Computer vision and robotics,
- Programming language compilers and hardware design,
- Database management systems (DBMS) parameters configuration.
To learn about the core principles of HyperMapper refer to the papers section at the bottom.
For any questions please contact Luigi Nardi.
HyperMapper is distributed under the MIT license. More information on the license can be found here.
Luigi Nardi, Assistant Professor, Lund University, and Research Scientist, Stanford University
Artur Souza, Ph.D. Student, Federal University of Minas Gerais
Bruno Bodin, Assistant Professor, National University of Singapore
If you use HyperMapper in scientific publications, we would appreciate citations to the following paper:
Practical design space exploration (MASCOTS 2019) - introducing HyperMapper principles and application to hardware design space exploration:
@inproceedings{nardi2019practical,
title={Practical design space exploration},
author={Nardi, Luigi and Koeplinger, David and Olukotun, Kunle},
booktitle={2019 IEEE 27th International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunication Systems (MASCOTS)},
pages={347--358},
year={2019},
organization={IEEE}
}Spatial: A Language and Compiler for Application Accelerators (PLDI 2018) - conference paper on the application of HyperMapper to the Spatial programming language:
@inproceedings{koeplinger2018spatial,
title={Spatial: a language and compiler for application accelerators},
author={Koeplinger, David and Feldman, Matthew and Prabhakar, Raghu and Zhang, Yaqi and Hadjis, Stefan and Fiszel, Ruben and Zhao, Tian and Nardi, Luigi and Pedram, Ardavan and Kozyrakis, Christos and others},
booktitle={Proceedings of the 39th ACM SIGPLAN Conference on Programming Language Design and Implementation},
pages={296--311},
year={2018},
organization={ACM}
}Algorithmic Performance-Accuracy Trade-off in 3D Vision Applications Using HyperMapper (iWAPT 2017) - workshop on the application of HyperMapper to ElasticFusion and KinectFusion computer vision applications:
@inproceedings{nardi2017algorithmic,
title={Algorithmic performance-accuracy trade-off in 3D vision applications using hypermapper},
author={Nardi, Luigi and Bodin, Bruno and Saeedi, Sajad and Vespa, Emanuele and Davison, Andrew J and Kelly, Paul HJ},
booktitle={Parallel and Distributed Processing Symposium Workshops (IPDPSW), 2017 IEEE International},
pages={1434--1443},
year={2017},
organization={IEEE}
}Integrating algorithmic parameters into benchmarking and design space exploration in 3D scene understanding (PACT 2016) - conference paper applying an early version of HyperMapper to 3D computer vision:
@inproceedings{bodin2016integrating,
title={Integrating algorithmic parameters into benchmarking and design space exploration in 3D scene understanding},
author={Bodin, Bruno and Nardi, Luigi and Zia, M Zeeshan and Wagstaff, Harry and Sreekar Shenoy, Govind and Emani, Murali and Mawer, John and Kotselidis, Christos and Nisbet, Andy and Lujan, Mikel and others},
booktitle={Proceedings of the 2016 International Conference on Parallel Architectures and Compilation},
pages={57--69},
year={2016},
organization={ACM}
}Application-oriented design space exploration for SLAM algorithms (ICRA 2017) - conference paper on the application of HyperMapper to robotics:
@inproceedings{saeedi2017application,
title={Application-oriented design space exploration for SLAM algorithms},
author={Saeedi, Sajad and Nardi, Luigi and Johns, Edward and Bodin, Bruno and Kelly, Paul HJ and Davison, Andrew J},
booktitle={Robotics and Automation (ICRA), 2017 IEEE International Conference on},
pages={5716--5723},
year={2017},
organization={IEEE}
}