Inspiration-Q develops quantum-inspired algorithms for mathematically hard problems in optimization, simulation and machine learning. Our technology is based on decades of progress in the understanding and modelization of complex quantum systems and quantum correlations. However, unlike quantum computing algorithms, which are often cited in the literature, quantum-inspired computing is a relatively new concept that needs some clarification. This article is devoted to answering some of the most fundamental questions and confusions we have observed around this topic.
- What is a quantum-inspired algorithm?
- Why are quantum-inspired solutions better than traditional ones?
- Will quantum computers make all classical computation obsolete?
- Will quantum-computers make quantum-inspired solutions obsolete?
- Are you against quantum-computing?
- Are quantum-inspired solutions compatible with quantum computers?
- Can we replace a quantum computer with a quantum-inspired technology?
- Do you just simulate a quantum computer? Is this like Qiskit or Cirq?
- Is quantum-inspired computation just a form of supercomputing?
What is a quantum-inspired algorithm?
The term quantum-inspired denotes that we are working with an algorithm that has its origins in the study of quantum systems or, more recently, in the field of quantum computation. However, despite their quantum origin, they are classical algorithms which run in ordinary computers and are developed with the usual Turing-complete languages (C, C++, Python, etc).
Some contexts where quantum-inspired techniques have been discovered are the following ones:
- Techniques such as Quantum Monte Carlo appeared long ago to study complex quantum systems in Condensed Matter Physics. The study of quantum optimization algorithms has motivated an improvement or simplification of these methods, arising to a family of quantum annealing techniques that can be used to efficiently address classical optimization problems.
- The study of quantum computing techniques, has led also to the development of new classical algorithms. The most notable case is Ewin Tang’s recommendation algorithm, which provided the same speed up of earlier quantum recommendation algorithms, making them obsolete!
- In the last 20 years, there have been great advances in the development of tensor network algorithms that are used to describe the strong correlations present in quantum computers. We now know that such techniques can be used to solve complex problems in machine learning or in numerical analysis, such as the anomaly detection task or the solution of partial differential equations.
Why are quantum-inspired solutions better than traditional ones?
Quantum-inspired algorithms benefit from the insight and the understanding of hard computational problems using quantum information frameworks.
- Quite often, a simple redefinition of the problem opens the door to innovative solutions with significant speedups.
- Other times, we can transform quantum algorithms into classical algorithms that provide heuristic speedups in many practical situations. We could not find those algorithms without understanding how quantum systems work.
- Yet other times, quantum-inspired solutions make use of state-of-the-art technology, such as tensor network methods, that were developed purely for quantum mechanical studies, but which have great potential in many other fields.
Will quantum computers make all classical computation obsolete?
Universal quantum computers are also universal quantum Turing machines. This means that, formally, a universal quantum computer can run two extreme types of programs
- It can run classical algorithms that work with bits much like an ordinary processor, albeit with some interesting properties, such as performing reversible computation.
- It can run quantum algorithms that work with qubits, quantum superpositions and entanglement, opening the door to applications such as factoring, efficient matrix inversion, quantum neural networks, etc.
People often conclude that quantum computers will therefore make all classical computers obsolete, and that classical techniques, such as quantum-inspired computation, will become an interim solution, soon to be abandoned. This is not true, for various reasons:
- Quantum computers are fragile. They need very sophisticated error correction techniques that are very expensive and require huge redundancies (100x to 1000x qubits to encode just one logical, error free qubit). It is therefore a waste of resources to run classical computations on existing quantum computers.
- Classical algorithms are not magically accelerated by running on a quantum computer. Quite on the contrary. Qubits are usually slower than ordinary bits in a classical computer, working at speeds of 10’s of MHz, vs. giga-Hertz clock speeds of silicon processors.
- Quantum algorithms do not always provide a consistent speedup over classical algorithms. For instance, for many practical problems, not those of extreme hardness, a quantum algorithm might require very little entanglement. In that case, it is preferrable to use a quantum-inspired algorithm that will run faster on ordinary hardware.
- We still do not have good enough quantum algorithms and realistic applications where quantum computers provide a substantial and future-proof speed-up. This is a very active research field that needs further development, and quantum-inspired solutions can help ensure whether those speed-ups are right or not.
Will quantum-computers make quantum-inspired solutions obsolete?
No. Quantum-inspired solutions provide a measurable advantage in many applications. As explained before, in those same problems the quantum computer might not have an exponential speedup, or when applied to the problem at hand, the quantum algorithm does not exploit the potential of the quantum computer — simply because the problem does not require it! In those situations a quantum-inspired algorithm is just a more scalable, more affordable approach.
Are you against quantum-computing?
No!!! Quite the contrary. Inspiration-Q is founded by a team of experts and researchers that work on the development of both quantum-inspired and quantum computing algorithms. Indeed, one of the greatest sources of inspiration for us is to understand what things a quantum computer can do well and why and use that information to create new classical solutions. We therefore believe in the coexistence of multiple models of computation.
Are quantum-inspired solutions compatible with quantum computers?
Because they are inspired by quantum algorithms, quantum-inspired solutions share conventions, problem definitions and most characteristics with the corresponding quantum computing algorithms. Hence, due to the “black-box” nature of both approachers, both solutions are largely interchangeable. In case of doubt, our team can help you develop future-proof solutions that benefit from both approaches.
Can we replace a quantum computer with a quantum-inspired technology?
No. Quantum algorithms provide exponential speedups in many applications, such as factoring, for which there are no known classical alternatives. In exceptional situations those speedups are provably unique, and cannot be reproduced by classical computation means.
Do you just simulate a quantum computer? Is this like Qiskit or Cirq?
No. Some quantum-inspired algorithms that we develop are inspired by quantum computing, but they are designed from scratch to create new, extremely efficient algorithms that do not imitate universal quantum computers. For that we use multiple techniques, from tensor network methods that can reproduce highly correlated data, to compressed explorations of solution spaces that optimize those correlations.
Is quantum-inspired computation just a form of supercomputing?
Yes and no. Quantum-inspired algorithms can run on individual computers, on big clusters, on SaaS architectures and on many different types of architectures and installations. Our own API is delivered as an AWS Lambda service that can be upgraded to AWS E3 images running on demand, depending on the service contract.