Faster analysis of relationships between large sets of data is possible to be done with quantum computers compared to classical computers according to a study by an international team. Also, the analysis is now possible for a wider range of data types than was expected earlier.
A “quantum linear system algorithm” was proposed by a study that was published late last week in American journal Physical Review Letters. The new algorithm is able to analyze data related to problems like as commodities pricing, social networks and chemical structures and is able to undertake quantum computation that can be used for artificial intelligence.
Research on artificial intelligence or quantum forms of machine learning have has been spurred on by the linear system algorithm that operates on a large matrix of data and was first propounded in 2009.
“There is a lot of computation involved in analyzing the matrix. When it gets beyond say 10,000 by 10,000 entries, it becomes hard for classical computers,” explained Zhao Zhikuan, the paper’s corresponding author from Singapore University of Technology and Design.
Application to a very specific type of problem is possible by the previous quantum algorithm of this kind, Zhao also said. This is due to the fact that with the rise in the number of elements in the matrix, there is a corresponding rapid increase in the number of computational steps. There is an eight-fold increase in the length of the calculation for every doubling of the size of the matrix.
Compared to both the classical and the previous quantum versions, a new algorithm was found to be faster by colleagues of Zhao in Singapore, Switzerland and the United Kingdom, he said.
The new system is dependent on a technique that is also known as quantum singular value estimation. This technique is able to deal with data that is not limited to “sparse” ones as had been required previously by the earlier versions.
Zhao said that while the relationships among the elements in sparse data is limited, for real-world data, this doe not often hold true.
Take for example the attempt of a trader to forecast the price of goods in the future. This new matrix is able to get hold of historical data related to movements of price over a certain period of time as well as information about features which might be impacting the prices such as the rates of currency exchange.
Then, by “inverting” the matrix, the new form of algorithm is able to calculate the strength of the correct relationship of each feature with another one. The resultant data can be made use of to understand the future price by the process of extrapolation.
But larger quantum computers would be needed to exhibit that the real quantum has an advantage compared to the classical algorithms.
According to estimates by Zhao, another three to five years would be required till such time that the hardware can be actually put to use for conducting any form of useful quantum computation that would also have some applications in artificial intelligence.
(Adapted from Xinhuanet.com)