Reduce Run_Time in hybrid CQM solver
I am solving 01 KnapSack problem for 500k items with the help of hybrid CQM solver of D Wave. And for comparison I solved same problem with CPLEX. Solution quality of CPLEX solver is better than D-Wave (Not Surprising) but what is surprising for me is that the Run_Time for D-Wave CQM solver (default 85 sec) is also higher than CPLEX solver (16 sec).
I know CQM solver have to sample the problem many times. But my question is Whats the point of using quantum computer to solve optimization problem if they give inferior solution in more run_time than Classical Solver.
If anyone knows how can I reduce the run_time for D-Wave then please suggest some
Comments
Hello,
Thank you for reaching out to us. It seems that you have already been provided the following information in a separate thread, but we wanted to follow up here as well to close the loop and answer your question in the community.
When running small problems (a 5 item knapsack problem) it is expected for classical solvers to outperform our hybrid solvers. The overhead of the cloud platform will increase our run time, and the problem is simple enough to run classically.
The Knapsack problem in its basic form is also a linear problem, which can be solved efficiently on classical solvers (such as CPLEX). D-Wave solvers are particularly performant for quadratic problems, with pairwise interactions between variables. Running a more advanced version of this problem (with multiple knapsacks, or interaction between items) will be a better way to showcase the advantage of quantum annealing.
When using a hybrid solver, the time_limit parameter is the only one you can access. The number of samples ran will be automatically set. If you run on the QPU directly you can select the number of samples to run. In both cases trial and error is the best approach to find the correct value. Increase the time_limit until there is no improvement to the solution quality.
Please sign in to leave a comment.