News Releases
Mitsubishi Electric's New Quantum Artificial Intelligence Technology Uses Automated Design to Realize Compact Inference ModelsCan be incorporated in first-ever application for terahertz imaging
FOR IMMEDIATE RELEASE No. 3560
TOKYO, December 2, 2022 - Mitsubishi Electric Corporation (TOKYO: 6503) announced today its development of a quantum artificial intelligence (AI) technology that automatically designs and optimizes inference models to downsize the scale of computation with quantum neural networks. The new quantum AI technology can be integrated with classical machine learning frameworks for diverse solutions.
Mitsubishi Electric has confirmed that the technology can be incorporated in the world's first1 application for terahertz (THz) imaging using ultra-high-frequency waves2 to perform non-destructive inspections by exploiting properties such as the high penetration of radio waves and the high directivity of light waves. It also can be used for indoor monitoring in which Wi-Fi signals are used to observe room environments for human movements. Other potential applications also are possible, such as compressed sensing to retrieve original data from mixed measurement data and biosignal processing for brain-computer interfaces.
Mitsubishi Electric's new quantum machine learning (QML) technology realizes compact inference models by fully exploiting the enormous capacity of quantum computers to express exponentially larger-state space with the number of quantum bits (qubits). In a hybrid combination of both quantum and classical AI, the technology can compensate for limitations of classical AI to achieve superior performance while significantly downsizing the scale of AI models, even when using limited data.
Rapidly developing quantum computers are expected to surpass classical computers by exploiting quantum physics to manipulate states of qubits in a highly parallel manner. Major breakthroughs are envisioned in data analysis, AI development, etc. for purposes such as large-scale optimizations and new material design. Classical machine learning technologies3 based on deep learning4—the cornerstone of current AI—have shown excellent performance but require costly computing resources and often do not achieve full potential if training data5 or resources are limited.
Mitsubishi Electric will partially introduce its quantum AI technology and related achievements at a tutorial session of the IEEE Global Communications Conference (GLOBECOM) 2022.
- 1According to Mitsubishi Electric's research as of December 2, 2022
- 2Electro-magnetic wave having a frequency around 0.1-10THz
- 3Machine learning designed to run on conventional computers
- 4Machine learning method to train computers for voice recognition, image identification, motion prediction, etc.
- 5Dataset containing signals and/or associated labels that AI models and machine learning algorithms can use to solve tasks
Note
Note that the press releases are accurate at the time of publication but may be subject to change without notice.