Foundations, Design and Implementation
Fred Glover, University of Colorado, Boulder, January 2018
Gary Kochenberger, University of Colorado, Denver, January 2018
Andy Badgett, Meta-Analytics, Inc., January 2018
Rajesh Chawla, Meta-Analytics, Inc., November 2018
Cesar Rego, University of Mississippi, February 2018
Jin-Kao Hao, University of Angers, February 2018
Zhipeng Lu, Huazhong University of Science & Technology, February 2018
Yang Wang, Northwestern Polytechnical University, February 2018
Development started March 2018
Alpha-QUBO 1.0 - May 2018 (initial experimental version; language C)
Alpha-QUBO 1.1 – July 2018 (First Web Inteface; additional languages and tools C, C++, C#, ASP.net core)
- Initial QUBO solver
- Tested on intel i7 4 core machine
Alpha-QUBO 1.2 - October 2018 (Improvements based on user feedback)
- Web Interface, Authentication, Upload of QUBO models, Download of solutions
- Amazon EC2 cloud deployment
- Tested on Amazon t2.medium instance
Initial Users and feedback
Gary Kochenberger, UC Denver, July 2018
Mark Lewis, Missouri Western State University, August 2018
Scott Pakin, LANL, October 2018
- Large file upload, Option to solve for minimization, model sharing between users
- Tested on Amazon t2.large instance
Alpha-QUBO 1.2 Current Features:
- Authentication: Users can sign up for an account, upload data to secured authenticated location.
- Web interface: Management of QUBO models, solutions and jobs.
- Large file upload: Files can be optionally compressed prior to upload to improve upload speed.
- Solution download: Multiple solutions for a model can be downloaded as a text file.
- Solve for Minimum or Maximum solution:
- Sharing of models: Users can share QUBO models with other users.
- QUBO model solution generation: Users can enter parameters and create a new QUBO solution. The solver runs on a server in the cloud.
- Scalable cloud architecture: All data stored is in the cloud and is available for processing on all available cloud machines.
Features in progress
- Multiple solutions for a single QUBO job run
- Double precision QUBO model values
- Visualization of objective function over time using web interface
- Advances in efficiency from innovations in algorithm design and software refinements.
- New data structures and software designs to effectively handle problems with larger dimensions.