Source: ATYUN AI platform
Singapore Center for Quantum Technology (CQT) researchers proposed a new algorithm for solving linear equations, the algorithm is faster than traditional quantum and previous versions, data types and is not limited.
Linear equations involving the problem of commodity prices, social networking and chemical structure.
Linear system algorithm for large data matrix. For example, for trying to predict the future price of the commodity traders, the matrix may capture data historical price movements and the possible impact of these price data features, such as currency exchange rates. The algorithms to calculate the intensity of each feature associated with another feature of the "inverted" matrix. This information can be used to infer the future.
Matrix analysis involves important computing. For a conventional computer, once the size of more than 10,000 multiplied by 10,000, it will become very difficult, because the number of calculation steps increases rapidly with the number of elements in the matrix. Doubling the size of each matrix, the calculated length is increased eightfold.
The first quantum algorithm for linear system is proposed by another group of researchers in 2009, launched a study of the quantum form of machine learning or artificial intelligence. As the "Nature" magazine article stated, quantum computers and the information may be compressed feature extracted from the data calculated and mapped to a bit or a quantum qubit, rather than the traditional calculations required in monotonous number crunching. The article says: "Quantum machine learning using the results of the algebraic operation, and make good use of the obtained data may be divided into a plurality of groups - - - is a core task handwriting and speech recognition software, another search. mode. "Thus, a large amount of information may be treated with a relatively small quantum.
2009 algorithm can better handle larger matrix, the index provides advantages over classical algorithm, but only if their data is called "sparse" because most of the elements in the matrix are zero. In these cases, the relationship between the elements is limited, and real-world data is often not the case.
The new algorithm is faster, there is no limitation on the type of data. As a rough guide, for 10,000 square matrix, the classical algorithm will take the order of trillions of computational steps, the first step of a quantum algorithm is about 10,000 steps, new quantum algorithms only 100 steps. This algorithm relies on the technique known as quantum estimated singular values.
CQT of Jansen (Zhikuan) Zhao, Anupam Prakash and his collaborators Leonard Wossnig issued a "quantum algorithm for linear systems," they made in the "Physical Review Letters" of February 2.
On a small-scale quantum computers, quantum linear system for early algorithm some proof of principle. Jansen and his colleagues hope to cooperate with an experimental group, their algorithm is proof of principle.
They also want to work to achieve the desired algorithm to conduct a comprehensive analysis to check the overhead costs that may exist.
Require greater quantum computer showed the real advantages over conventional algorithms. Jansen said: "We may be in the next three to five years, using hardware manufactured by laboratory personnel to perform meaningful quantum computing and applied artificial intelligence."
This switched ATYUN artificial intelligence media platforms, the original link: new quantum algorithm for linear systems can accelerate the speed of machine learning
more recommendations
Why Swift will be the next depth study of a favorite?
Alpha's new features: integration of monitoring training
PULP Dronet nano-UAVs weighing only 27 grams, with a strong autonomous navigation capability
