As a user of the OptimizationSolver for QUBO problems, I want the solver to chose the most efficient connectivity mode for synaptic connections to get the best EDP performance and scale the workloads to the maximum allowed by the hardware.
Conditions of Satisfaction
The solver computes the sparsity of the QUBO matrix and decides whether to implement synaptic connections as Dense or Sparse processes.
Acceptance Tests
For QUBO problems with sparsity below 50%, The model of the OptimizationSovler implements the CostMinimizer synapses as Dense processes.
For QUBO problems with sparsity above or equalt to 50%, The model of the OptimizationSovler implements the CostMinimizer synapses as Sparse processes
The ideal choice of sparse vs. dense does not only depend on sparsity but also idx and weight bits.
For high sparsity (e.g. 0.95, few zeros) we want to use Dense and for low sparsity (e.g. 0.05, many zeros) we want to use Sparse. Not the other way around.
User Story
As a user of the OptimizationSolver for QUBO problems, I want the solver to chose the most efficient connectivity mode for synaptic connections to get the best EDP performance and scale the workloads to the maximum allowed by the hardware.
Conditions of Satisfaction
Acceptance Tests
Lava version:
I'm submitting a ...
Other information: