• home_icon HOME
  • Research
  • Highlights
Quantum AI algorithm surpasses AI technology

Quantum artificial intelligence is a future technology that addresses the limitations of current artificial intelligence technology. The team, led by Professor June-Koo Rhee, developed an algorithm that significantly improves artificial intelligence technology.

Artificial intelligence relies on a kernel function to represent similarity between data sets. Supercomputers used in complicated learning are approaching physical limits in terms of space and energy. On the other hand, quantum artificial intelligence is capable of performing complex calculations with less burden. However, quantum artificial intelligence processes information in a completely different way compared to conventional computers. Few studies exist on this topic despite the need for new algorithms. The team developed a quantum machine learning algorithm to handle complex data, hinting at the possibility of quantum artificial intelligence surpassing current artificial intelligence technology.

Similarity between quantum data sets demonstrated using the IBM Q cloud

The team, led by Professor June-Koo Rhee, developed a quantum machine learning algorithm that achieves exponentially higher computational efficiency relative to the size of the training data by utilizing the advantages of quantum computing. The algorithm transfers data existing in lowdimensional input space to high-dimensional feature space, calculates the weighted power sum of the fidelities of quantum data in quantum parallel, and efficiently determines the classification of test data. One advantage offered by the algorithm is that the computational complexity of the quantum circuit increases very slowly due to logarithmic dependence.

While quantum circuits known to data were only capable of linear classification, the proposed algorithm opens up the possibility of complex nonlinear kernel classification based on the systematic design of quantum circuits. Various quantum kernels can now be effectively modeled, as optimal kernels vary with input data in kernel-based machine learning. The team also succeeded in demonstrating the superior performance of quantum kernel-based machine learning using the IBM Q cloud with five superconducting qubits in the quantum circuit.

“Quantum computers will surpass supercomputers in 3 to 5 years”

Professor June-Koo Rhee emphasized the need for follow-up research, saying, “In three to five years, quantum computers will surpass supercomputers in performance. We can expect quantum computers to have an exponential increase in computing capacity, allowing them to replace supercomputers. Thus, research in this area is essential.”

If high-performance NISQ computing technology capable of controlling hundreds of cubits at 99.9% accuracy is secured in the next few years, the proposed algorithm will be actively utilized as a quantum machine learning algorithm for pattern recognition of complex nonlinear data, extending beyond kernel-based quantum supervised machine learning. The study was conducted jointly with research teams from Germany and South Africa, and the results were published in npj Quantum Information in May 2020.

Prof. June-Koo Rhee
2020 KI Annual Report

KAIST 291 Daehak-ro, Yuseong-gu, Daejeon (34141)
T : +82-42-350-2381~2384
F : +82-42-350-2080
Copyright (C) 2015. KAIST Institute