CEQT Center of Excellence in Quantum Technology

Research › Algorithm design

Pillar 1 · Research

Algorithm design

Complexity theory, QUBO formulation, adiabatic and annealing algorithms, hybrid classical–quantum methods, quantum machine learning and quantum-inspired classical algorithms.

Turning hard problems into forms a quantum machine can attack.

A quantum computer does not solve a problem because the problem is hard. It solves a problem because the problem has been put into a form the machine’s physics can express. That translation step — from a network, a portfolio, a schedule or a threat model into a Hamiltonian or a circuit — is where most of the intellectual work lives, and it is what this pillar does.

Much of our work goes through quadratic unconstrained binary optimisation (QUBO), because QUBO is the common currency of annealing hardware and of several hybrid solvers. Constructing a good QUBO is not mechanical: constraints have to be turned into penalties, penalties have to be weighted so that they neither dominate nor vanish, and the resulting problem has to stay inside the connectivity the hardware actually offers. New construction techniques developed for one domain are often formulated so that they can be carried over to others.

We are equally interested in the negative result. A method that turns out not to beat a good classical heuristic is worth publishing, and worth saying so on a page like this one.

From a problem to something a quantum machine can attackthe classical baseline runs the whole way alongside
How a problem becomes something a quantum machine can attackA problem is written as a cost function, then recast in QUBO or Ising form — binary variables and a coupling matrix — and handed to a solver. The centre's work is concentrated on the recasting step. Running underneath the whole chain, the same instance is also solved the established classical way, which is the number every quantum result is compared against.The problema routing, schedulingor key-search instanceA cost functionwritten so that lowermeans betterQUBO / Ising formbinary variables anda coupling matrixA solverannealer, gate machine,or classical heuristicThe classical baselinethe same instance, solved the established way —the number every quantum result is measured againstWhere CEQT works
Over the barrier, or through itthe distinction quantum annealing turns on
Leaving a local minimum: over the barrier, or through it A double well, the shallower one on the left. To reach the lower well on the right a classical search must climb the whole height of the barrier between them, and the rate at which that happens falls away exponentially with that height divided by the temperature. A quantum state instead crosses at its own energy, straight through the barrier; the amplitude for doing so falls away with the barrier's height and its width together. Tunnelling therefore helps on barriers that are tall and narrow, and not on ones that are low and broad. Through — quantumcrosses at its own energy; transmission ~ e−(2/ℏ)∫√(2m(V−E))dxOver — classicalmust find ΔE; rate ~ e−ΔE/kT ΔE w, at this energy trapped here the answer

When tunnelling actually helps

Height alone governs the classical route, so a low broad barrier is easy to climb and a tall one is not. Tunnelling depends on height and width together, which is why it is the better route through barriers that are tall and narrow, and the worse one through barriers that are low and broad.

That is a statement about barriers, not about problems. Whether a real instance has the kind of landscape where this converts into an advantage over a good classical heuristic is an open question, and it is the question this pillar works on. The classical baseline on every application page is there for exactly this reason.

Methods we use

  • QUBO and Ising formulation — constraint encoding, penalty weighting, embedding onto hardware connectivity.
  • Adiabatic quantum computation and quantum annealing — including probability-boosting techniques for combinatorial optimisation.
  • Hybrid classical–quantum algorithms — decomposition, warm starts, and classical post-processing of quantum samples.
  • Complexity analysis — where a problem sits, and what that implies about any speed-up claim.
  • Quantum machine learning — variational models, kernel methods, and the question of when the quantum part is doing the work.
  • Quantum-inspired classical algorithms — methods derived from quantum structure that run on ordinary hardware.

Problems we apply them to

  • Attack and defence strategy on layered networks.
  • Portfolio and risk optimisation under realistic constraints.
  • Routing, scheduling and assignment problems.
  • Feature selection and model search in machine learning.
  • Resource allocation in energy systems.

Active in: Cybersecurity & defence · Finance & econometrics · Logistics & optimisation · Energy & smart grids · Machine learning · Complex systems

Current work

  • In preparation

    A chained partial-sum formulation for knapsack QUBOs

    Replacing one dense budget constraint with a chain of smaller equalities over intermediate registers: exact, and far easier to minor-embed on D-Wave Pegasus and Zephyr hardware. What it does →

  • On-going

    Research Enhancement in Quantum Technology

    Strengthening the centre’s algorithmic and computational capability across the pillars.

  • Proposed

    Advanced quantum computing research with IBM-Q

    Gate-model algorithm development on production quantum hardware.

  • Proposed

    Human development for excellence in quantum computing, for collaboration with AI technology

    Building the researcher pipeline at the intersection of quantum computation and machine learning.

People in this pillar

Assoc. Prof. Dr. Sanpawat Kantabutra

Complexity theory and quantum algorithms · Director

sanpawat.k@gmail.com

Assoc. Prof. Dr. Chulin Likasiri

Applied mathematics and quantum computing

Assoc. Prof. Dr. Wattana Jindaluang

Approximation algorithms and graph theory

wattana.jinda@cmu.ac.th

Everyone at the centre →

Selected publications

  • Kantabutra, S. “Adiabatic Quantum Computation for Cyber Attack and Defense Strategies.” In New Trends in Computer Technologies and Applications, ICS 2022, Communications in Computer and Information Science vol. 1723, Springer, Singapore. doi:10.1007/978-981-19-9582-8_9
  • “Hybrid classical quantum computation for cybersecurity strategies in a layered cybersecurity model.” The Journal of Supercomputing 81, 1179 (2025). doi:10.1007/s11227-025-07662-4
  • “Probability-boosting technique for combinatorial optimization.” PeerJ Computer Science 10:e2499 (2024). doi:10.7717/peerj-cs.2499

Browse all 215 publications →

See also

Annealing and gate-model hardware access is described under Software & platforms. The complexity results that underpin this work sit in Foundations. For the cold-atom platforms themselves, see RCQT.

Last updated 14 September 2026.