insider brief
- Researchers report that they may have found a way to bridge the “reality gap,” the “crazy golf-like” difference between predicted and observed behavior from quantum devices.
- “Physically-based” machine learning approaches may be able to indirectly infer the characteristics of a fault.
- The team was led by scientists from the University of Oxford.
Press Release — Research led by the University of Oxford harnesses the power of machine learning to overcome key challenges affecting quantum devices. This discovery reveals for the first time how to bridge the “reality gap,” or the difference between predicted and observed behavior from a quantum device. The results were announced below. Physical Review X.
Quantum computing has the potential to greatly enhance a wealth of applications, from climate modeling and financial forecasting to drug discovery and artificial intelligence. However, this will require effective ways to scale and combine individual quantum devices (also known as qubits). The main barrier to this is the inherent variability. In other words, seemingly identical units behave differently.
Functional variations are presumed to be caused by nanoscale imperfections in the materials that make up quantum devices. Because there is no way to directly measure these, simulations cannot capture these internal disturbances, creating a gap between predicted and observed results.
To address this, the research group used a “physics-based” machine learning approach to indirectly infer the characteristics of these disorders. This is based on how internal disturbances affect the flow of electrons through the device.
Lead researcher Associate Professor Natalia Ares, from the University of Oxford's School of Engineering, said: “For example, when we play 'crazy golf', the ball goes into a tunnel and travels at a speed or direction that does not match our expectations. There is a possibility of getting out.” . But with a few more shots, a crazy golf simulator, and some machine learning, we might get better at predicting ball movement and close the gap with reality. ”
The researchers measured the output current of individual quantum dot devices at various voltage settings. The data was input into a simulation to calculate the difference between the measured and theoretical currents in the absence of internal faults. By measuring the current at many different voltage settings, the simulation was constrained to find an internal fault configuration that could explain the measurements at all voltage settings. This approach used a combination of mathematical and statistical approaches and deep learning.
Professor Ares added: “In a crazy golf analogy, this is equivalent to installing a series of sensors along a tunnel, allowing you to measure the speed of the ball at different points.” Although you can't see inside the tunnel yet, you can use this data to more accurately predict how the ball will behave when you take a shot. ”
The new model not only found a suitable internal fault profile to explain the measured current values, but was also able to accurately predict the required voltage settings for a given device operating regime.
Importantly, this model provides a new way to quantify variation between quantum devices. This allows us to more accurately predict how the devices will behave and could also help us design the best materials for quantum devices. It could inform compensation approaches to reduce the undesirable effects of material defects in quantum devices.
Co-author David Craig, a PhD student in the School of Materials at the University of Oxford, said: 'Although we cannot directly observe black holes, in the same way that we can infer their existence from their effects on surrounding matter, We have been using simple measurements as a tool,” he added. A proxy for internal fluctuations in nanoscale quantum devices. Although real devices have more complexities than models can capture, our work demonstrated the utility of using physics-aware machine learning to narrow the reality gap. . ”
Check out the latest information for more market insights. Click here for quantum computing news.
