30.03.2008
Researchers from Siemens Corporate Technology (CT) and Munich Technical University have achieved the world’s first experimental implementation of an artificial neural network on a simple quantum computer. This development is a step in the direction of realizing a practical quantum computer. The simulation of pattern recognition conducted by the CT researchers was evaluated in a magnetic resonance spectrometer (NMR) by scientists from the university. The system with two "qubits" confirmed all results.
The simultaneous progress of miniaturization and increased performance of conventional computers will come up against physical limits in a few years’ time. That’s why experts expect computers in the future will calculate with qubits instead of bits. While one bit can represent only 0 or 1, qubits can simultaneously take on different states and also be in superposition with other qubits, due to the curious nature of the quantum world. In addition to making faster computing possible, it is hoped this also will result in reduced energy consumption, which is already very high for today’s mainframe computers.
The experts from CT achieved their quantum computer research results using special neural networks that are suitable for recognizing patterns. The pattern used is composed of dots, and each dot can assume two colors, which are depicted by means of qubits. Using an algorithm they developed, the researchers can predict how a real quantum processor would behave if confronted with a new color pattern. The simulation compares this pattern with patterns stored in the memory and indicates the degree of similarity.
The actual experiments were conducted by the Munich Technical University researchers in an NMR spectrometer. They used a solution of Sodium formiate at room temperature, which contains one carbon and one hydrogen atom. In strong magnetic fields, each of the two particles forms one qubit. The measured signals of the actual quantum computer corresponded exactly to the signals that had previously been calculated. The researchers thus showed that their algorithm delivers correct results for a quantum computer in actual practice.
Faster pattern recognition – which is the long-term goal of the development – could be used in a broad spectrum of applications at Siemens, from medical technology to automation or energy technology. A quantum computer would be very well-suited to complex problems such as identifying genetic sequences. The researchers’ vision is a hybrid processor that would work with conventional technology and quantum mechanical methods. While conventional chips would be able to handle most operations, a quantum processor would take over certain tasks.
