Brief Summary
Dr. Pascal Hoffman from Fraunhofer ITWM discusses the state of quantum optimization, focusing on practical applications. He explains optimization problems, compares classical and quantum solutions, and suggests areas where quantum computing may excel, such as multi-objective and stochastic optimization. The talk concludes with a Q&A session, addressing the potential of quantum computers in various industries and the challenges in proving quantum advantage.
- Optimization problems consist of optimizing a function with constraints and decision variables.
- Quantum computing has the potential to solve NP-hard optimization problems, but classical methods are strong.
- Quantum algorithms like quantum annealing, VQE, and QAOA show promise, but their performance varies on current hardware.
- Multi-objective, robust, and multi-stage optimization problems are more suitable for quantum computing.
- Hybrid algorithms that combine quantum and classical computing are a promising approach.
Introduction to Fraunhofer and Quantum Computing
Dr. Pascal Hoffman is introduced as the research coordinator for quantum computing activities at Fraunhofer ITWM. Fraunhofer is a leading applied research organization with a mission to transition fundamental research into ready-to-use solutions for industry. Their work spans various sectors, including automotive, finance, energy, material science, and healthcare. The organization aims to make quantum industry-ready by accelerating its utility and raising awareness among industrial partners, focusing on finance and material science use cases, integrating both classical and high-performance computing hardware.
Optimization Problems: Classical vs. Quantum
Optimization problems involve minimizing or maximizing an objective function subject to equality and inequality constraints, with decision variables that can be continuous, integer, or mixed. Examples include the traveling salesperson problem and unit commitment in the energy industry. Although these problems are NP-hard, classical solvers can solve them quickly for smaller instances. Quantum computing is often proposed for NP-hard problems, but classical methods set a high bar for quantum utility or advantage.
Quantum Optimization Algorithms
Several quantum algorithms exist for optimization, including quantum annealing, which uses adiabatic evolution and quantum tunneling. Variational Quantum Eigensolver (VQE) uses parameterized quantum circuits optimized by classical solvers. The Quantum Approximate Optimization Algorithm (QAOA) is a specialized VQE that mimics quantum annealing. Grover Adaptive Search marks items with objective functions below a threshold and iteratively lowers the threshold. VQE and QAOA are promising as they can be executed on current hardware and leave room for fault-tolerant quantum computing.
Performance of Quantum Optimization Methods
Grover adaptive search performs poorly on current quantum hardware due to lengthy quantum circuits. Quantum annealing, used by D-Wave, has shown major updates, with D-Wave proposing quantum supremacy and use cases in drug discovery and navigation systems. However, experiments show that classical solvers can outperform quantum computers on smaller instances of problems like unit commitment. The current state of quantum computing is between having no utility and achieving supremacy.
Suitable Problems for Quantum Computing
Quantum computing is more suitable for problems that are harder to solve classically and feature real-world complexities. Multi-objective optimization problems, like portfolio optimization, involve conflicting objectives and require finding Pareto optimal solutions. Robust and stochastic optimization deals with uncertain parameters and random variables, aiming to find solutions that perform well across different scenarios. Multi-stage and multi-level problems, such as transportation planning, can potentially be solved holistically with quantum computing.
Q&A: Quantum Computing Applications and Hardware
The Q&A session addresses several key points:
- NP-hard problems are not always hard in practice due to advanced classical optimization techniques.
- Quantum computing may succeed in multi-objective problems where commercial solvers are lacking.
- Quantum algorithm plus quantum hardware is the benchmark against classical systems.
- Stochastic optimization problems, such as those in portfolio theory, are promising for quantum computers due to their stochastic nature.
- D-Wave's proposed quantum advantage requires expertise and a perfectly calibrated system, with problems closely fitted to its architecture.
- Hybrid algorithms are favored, especially in the current era of Noisy Intermediate-Scale Quantum (NISQ) devices.
- Established companies can integrate quantum computing through cloud access and supercomputing centers.
- Quantum advantage is less likely in single-objective portfolio optimization compared to multi-objective or uncertain parameter scenarios.
- D-Wave is well-suited for quadratic unconstrained binary optimization, while superconducting and ion trap qubits are interesting for universal quantum computers.
- Fraunhofer is exploring quantum computing for drug detection and quantum sensing in healthcare, but not specifically for optimization in that industry.
