Quantum Computer First Realizes Generative Confrontation Algorithms and Tsinghua University Team Leads the Completion

Home > Sci-Tech

Quantum Computer First Realizes Generative Confrontation Algorithms and Tsinghua University Team Leads the Completion

2019-01-31 10:25:28 469 ℃
Machine learning is becoming more and more powerful, thanks largely to an algorithm that confronts AI with each other. This kind of algorithm is called Generating Countermeasure Network (GAN), which can be used to create works of art, crack encryption, and generate realistic human and animal pictures.

Recently, researchers have combined the generation of antagonistic networks with quantum computing. Professor Sun Luyan and his colleagues at Tsinghua University's School of Cross-Information have used quantum circuits to implement generative countermeasure networks (GAN).

Figure This paper (Source: Science Advance)

Generating antagonism network is composed of two parts of neural networks, namely, generating model and discriminant model. As the name implies, the generation model is used to generate data, for example, he can generate a face image. Then the real data and the "false" data generated by the model are provided with the discriminant model to distinguish the true and false data. In this way, after the two neural networks are confronted with each other for a certain number of rounds, the generated model can generate the data that the discriminant model can not distinguish between true and false.

Scientists have realized quantum generation antagonism network through quantum generation model and discriminant model. Unlike classical computers that represent only 0 or 1 bits, a quantum computer can use a mixture of 0 and 1 bits.

Scientists tried to train quantum generation models to simulate the quantum data generated by microwave resonators until the discriminant model could not distinguish the true from the false. < p > < img SRC = "/ 1ydzximg / 0LDJYhgJY"/>

< p > Professor Deng Dongling, School of Cross-Information, Tsinghua University, said that in theory, quantum computers have absolute speed advantages over classical computers on specific issues, such as decomposition of large numbers. But according to Professor Sun Luyan, due to the current technical constraints, quantum computers can not achieve such an ideal speed.

Scientists believe that the generation antagonism network on quantum computers will have a speed advantage over classical computers, but they need to further verify this. Professor Sun Luyan said that this would be a landmark work in quantum machine learning.

Scientists generally believe that at least 50 qubits are needed to achieve "quantum hegemony", but this time their team used a single qubit system. Professor Deng Dongling also said that the road is blocked and long, and the line is coming.