Explore the cutting edge of quantum machine learning with quantum reservoir computing (QRC). This article details Keio University and Mitsubishi Chemical’s research using QRC to model and predict the behavior of a pliable soft robot.
IBM Quantum Innovation Centers
One of the first IBM Quantum Hubs in the world, now known as Quantum Innovation Centers (QICs), was Keio University in 2017. There are already more than 40 QICs worldwide. To propel developments in quantum computing, QICs leverage IBM Quantum’s skills and experience. By attracting participants to participate in collaborative research projects, these global hubs foster the development of a global quantum ecosystem, spearhead innovative quantum research, and create quantum learning communities.
As a QIC, Keio University collaborates with top Japanese industry players to create novel quantum applications and algorithms. The university’s continuous connection with Mitsubishi Chemical, a global leader in materials science research and development, is one example of these alliances. academics from the two organisations collaborated with academics from the University of Tokyo, the University of Arizona, and the University of New South Wales in 2023 to conduct a utility-scale experiment using an IBM Quantum device to implement a suggested quantum reservoir computing technique. An ongoing research project that is still going strong was started by this study.
Bringing utility-scale quantum computation to reservoir computing
A machine learning (ML) approach called reservoir computing (RC) seeks to lower the training overhead of more popular ML techniques like neural networks and generative adversarial networks. In this sense, a reservoir is a computational resource that has the ability to perform mathematical transformations on data from an input system, allowing for the manipulation of sizable datasets while maintaining the connections between individual data points.
Data from the input system is first sent to the reservoir by researchers in a standard reservoir computing experiment. Researchers will employ post-processing techniques to extract the necessary answers from the reservoir’s modified data output. The linear regression model, an ML model for characterising the connection between variables, is a common option for this post-processing. Researchers may use their linear regression model to create a time series that forecasts the behaviour of the input system after training it on modified output data produced by the reservoir.
A branch of RC called quantum reservoir computing (QRC) makes use of quantum processors as its reservoir. The processing of high-dimensional data is a natural fit for quantum computers, which might eventually outperform traditional systems in terms of computational capacity.
Mitsubishi Chemical, Keio University, and other research partners are investigating how quantum reservoir computing could support the understanding of complex natural systems. The goal of their 2023 experiment was to develop a quantum reservoir computing model that could forecast the noisy, non-linear motions of a “soft robot,” a pliable device whose motions are controlled by applying air pressure.
Developing new Quantum Reservoir Computing methods
To start the experiment, the researchers turned the robot’s movement data into quantum input states the IBM quantum reservoir could read. The reservoir received those inputs. After applying a series of random gates to the input states, the reservoir outputs signals that have been modified. After that, the researchers use linear regression to post-process these output data. A time series that should forecast the robots’ movements is the end outcome. Using benchmarking techniques, the researchers assess the accuracy of this prediction by contrasting it with real data.
Since measurement occurs at the conclusion of a quantum circuit in the majority of quantum reservoir computing techniques, you must repeatedly set up and execute the system for every qubit at every time step. This can raise the time required to complete the experiment and have a detrimental effect on the time series’ accuracy. Overcoming these obstacles with “repeated measurements” was one of the main objectives of the Keio University and Mitsubishi Chemical project.
They add more qubits to the system and measure these additional qubits repeatedly rather than setting up and operating the system at each time step. By using this technique, the researchers may gather the time series all at once, which results in a more accurate time series and less time spent running the circuits.
On IBM Quantum processors with up to 120 qubits, the researchers conducted practical demonstrations of their quantum reservoir computing system. In comparison to conventional QRC approaches, they discovered that their repeated measurements methodology produced greater accuracy and a noticeably lower execution time. Their first findings indicate that it should be able to further accelerate the calculation.
It will need a lot more study in the fields of RC and quantum reservoir computing before these techniques can provide meaningful solutions to issues. But according to the researchers, their utility-scale studies could already surpass traditional simulation techniques. They intend to investigate quantum reservoir computing in the future for challenging nonlinear issues like financial risk modelling.
How Quantum Innovation Centres benefit enterprise research organisations
One example of how businesses might gain from collaborating with partners like those found throughout the IBM network of Quantum Innovation Centers is the partnership between Keio University and Mitsubishi Chemical. Through these partnerships, professors and students who are not only highly skilled in quantum computing but also skilled at instructing other researchers in challenging topics may help enterprise researchers develop advanced quantum abilities.
Not just Mitsubishi Chemical, but other multinational corporations are also reaping these advantages. In addition to their collaboration with Mistubishi Chemical, Keio University is working with corporate R&D teams from top businesses in a variety of sectors and possible quantum use cases to explore fascinating areas of quantum applications and algorithm development. These collaborations indicate how industrial research trials carried out in collaboration with universities may open the door to worthwhile real-world applications and highlight the crucial role QICs can play in supporting corporate explorations of interesting quantum use cases.