Quantum Computing
Introduction to the world of Quantum Computing
Quantum computing leverages the fact that microscopic quantum mechanical variables can exist in multiple states simultaneously. This property, in principle, enables the parallel execution of computations that would otherwise need to be performed sequentially on classical computers. As a result, certain complex problems—those currently considered practically unsolvable—may become tractable, provided that suitable hardware and corresponding quantum algorithms can be developed. At present, however, quantum computing also presents several challenges. Quantum effects are typically observable only at extremely low temperatures, requiring intensively cooled and therefore expensive specialized hardware. Furthermore, the preparation of input states and the accurate readout of computational results are often non-trivial. The range of feasible computations is also significantly limited by architectural constraints of current hardware implementations—particularly regarding the number of qubits (quantum processors) and the permissible interactions between them.
​There is currently no universally accepted hardware standard for quantum computing. Several types of quantum hardware are being actively developed and investigated in both academic and industrial contexts. These include:​ superconducting quantum computers, trapped ion quantum computers, silicon-based quantum computers, photonic quantum computers and neutral atom quantum computers. ​Each of these platforms exhibits distinct advantages and limitations. Photonic quantum computing, for instance, is currently considered relatively inexpensive to implement because it does not require extensive cooling but is generally restricted to linear computations, although significant efforts are being made to expand its capabilities. At present, neutral atom quantum computers appear to offer the most promising balance between implementation cost and computational versatility, although this assessment may evolve rapidly.​
General-purpose quantum computers—devices that can be programmed to solve arbitrary problems of arbitrary size—do not yet exist. Despite optimistic claims by some manufacturers, the development of such systems is not expected within the next ten years. Only a limited number of quantum algorithms can currently be executed on existing hardware, and these are typically crafted as hardware-specific use cases. For example, Shor’s algorithm for factoring large numbers cannot yet be implemented at a practical scale, meaning that widely used cryptographic systems remain secure. Quantum Annealing (QA) is the only form of quantum computing that has already been commercially deployed for several years. It therefore represents not only a near-term but also a present-day quantum technology. Quantum Annealing does not aim to provide general-purpose computation. Instead, it is designed to solve a specific class of optimization problems. ​For simple optimization tasks, the optimum can be found easily. However, for heterogeneous optimization tasks, identifying the optimal configuration becomes computationally intractable for classical systems.
Quantum Machine Learning
A relatively recent area of research in the theory of information-processing systems is Quantum Machine Learning (QML). The central aim of QML is to design quantum analogues of classical machine learning architectures. Once realized, this development could significantly shift the trajectory of quantum computing—from general-purpose programmable devices to learning quantum systems. Within QML, one subfield of particular interest involves Quantum Boltzmann Machines (QBMs). These systems exhibit two notable features relative to quantum versions of conventional deep learning models. First, they are mathematically very similar to Quantum Annealers, a similarity that can be exploited in both design and implementation. Second, recent theoretical advances suggest that classical Boltzmann Machines may soon become interpretable. It is reasonable to expect that this interpretability will extend to their quantum counterparts as well.
Saddle Point's involvement
Why Saddle Point Science Europe? Although our current focus is on mathematical methods for medical data analysis, we realized early that our background, knowledge, and experience would be ideally suited for expanding into the development and implementation of innovative algorithms for Quantum Computing and Quantum Machine Learning (QML).
​
Developments in quantum computing are moving so fast that it will be impossible to become commercially competitive without early adoption. Japan is currently investing heavily in quantum computing, and quantum machine learning (QML) is expected to become the new focus. Developing algorithms for QML requires the ability to mathematically predict the evolution of quantum hardware, which is very similar to Quantum Annealers (QA). This type of mathematical prediction for QA is one of our company's intended priorities (and a topic on which we have already published), and our strong and extensive relationships with Japanese researchers increase our chances of making QML algorithms a commercial success in the future.
​
We have chosen to contribute to the following research themes:
-
Algorithms for optimizing Quantum Annealing dynamics - to enable better control of quantum annealers
-
Algorithms for Quantum Machine Learning at nonzero temperatures - to make quantum machine learning practical
-
Develop Interpretable Quantum Boltzmann Machines - to make quantum machine learning safe
