devdigest.today
#quantumoptimization #wireless #iot #networks #transferlearning | Mykola Maksymenko
The path toward useful quantum computing is not one breakthrough. It is a continuous process of innovation in inventing new algorithms, testing them on real problems, and running them on real quantum hardware.
Our research team at Haiqu is committed to this challenge. In a recent work that will be presented this week at the International Conference on Computer Communications and Networks, our team developed a new #QuantumOptimization approach for solving the connected dominating set problem, applied it to a practical network-design challenge, and tested it at scale on a QPU.
#Wireless mesh, #IoT, and ad-hoc #networks often need a small subset of devices to relay traffic and keep the full network connected. The challenge is to make this subset as small as possible without losing network coverage or connectivity.
Instead of asking a quantum computer to find the best answer in one attempt, our algorithm starts from a candidate solution, explores alternatives around it, and iteratively searches for improvements.
We implemented the approach on 73 qubits and ran it on an IBM Quantum Kingston quantum processor. To make the circuits executable on the hardware, we used Haiqu’s hardware-aware optimization, reducing the two-qubit gate count from 6,648 to 2,002 and the circuit depth from 980 to 335.
The method consistently outperformed the standard LR-QAOA quantum baseline and produced smaller virtual backbones on average. It can also use #transferlearning to carry parameters from smaller network instances to the larger problem, reducing the amount of costly tuning required directly on quantum hardware.
See Maxence Grandadam present this work at the International Conference on Computer Communications and Networks this Thursday, during the Quantum Machine Learning session.
