The researchers claim their circuits outperformed all baselines including circuits hand-designed by humans and those made using other computational methods.
MIT researchers claim to have developed a technique to make quantum circuits more resilient to noise, which can help boost performance and reduce errors in quantum computers.
The team said noise is a key challenge that holds back the field of quantum computing, as it can lead to higher error rates than classical computers. It is caused by imperfect control signals, interference from the environment, and unwanted interactions between qubits.
Performing computations on a quantum computer involves a quantum circuit, which is a series of operations called quantum gates. These gates change the quantum states of certain qubits in order to perform calculations but they also introduce noise, which can affect the performance.
In their study, the researchers said previous work for mitigating noise has primarily focused on gate-level or pulse-level noise-adaptive compilation. Instead of this method, the team focused on making the quantum circuit itself noise resilient.
The researchers said they created a framework that identifies the most robust quantum circuit for a particular computing task and then generates a mapping pattern.
They claim this method – called QuantumNAS (noise adaptive search) – is less computationally intensive than other search methods and can identify quantum circuits that improve the accuracy of machine learning tasks.
Senior author and associate professor, Song Han, said: “The key idea here is that, without this technique, we have to sample each individual quantum circuit architecture and mapping scenario in the design space, train them, evaluate them and if it is not good we have to throw it away and start over.
“But using this method, we can obtain many different circuits and mapping strategies at once with no need for many times of training.”
To construct a quantum circuit that is noise resilient, the team focused on variational quantum circuits, which use quantum gates with trainable parameters that can learn a machine learning or quantum chemistry task.
“With so many different choices, the design space is extremely large,” lead author Hanrui Wang said. “The challenge is how to design a good circuit architecture. With QuantumNAS, we want to design that architecture so it can be very robust to noise.”
In order to identify the quantum circuit that contains the ideal number of parameters, the researchers created a ‘SuperCircuit’, which contains all the possible parameterised quantum gates in the design space.
After training this circuit once, all other candidate circuits in the design space are subsets of this SuperCircuit, which means they inherit corresponding parameters that have already been trained.
The team then used this trained SuperCircuit to search for circuit architecture that has a high resistance to noise. Once they found and trained these circuits, they removed quantum gates that were causing noise and not contributing to performance. They then deployed these circuits in real machines.
The researchers said their circuits outperformed all the baselines when tested on real quantum devices, including circuits hand-designed by humans and others made using other computational methods.
“For QML tasks, QuantumNAS is the first to demonstrate over 95pc two-class, 85pc four-class, and 32pc 10-class classification accuracy on real quantum computers,” the study said.
The researchers have created an open-source library called TorchQuantum, that contains information about their projects, tutorials and tools that can be used by other research groups. The team hopes this will encourage further work in this area.
The research was supported by the US National Science Foundation, the MIT-IBM Watson AI Lab, the Qualcomm Innovation Fellowship and the US Department of Energy.
10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.