Google has unveiled TensorFlow Quantum, a quantum machine learning library, in collaboration with the University of Waterloo, Volkswagen, and research institute X(X). TensorFlow Quantum is a tool for modeling quantum structures by combining quantum computing and the machine learning research community.
TensorFlow Quantum is an integration of the existing TensorFlow library with Cirq, a framework developed by Google for NISQ, a quantum processor with quantum noise. It is realizing quantum computing that is compatible with the existing TensorFlow API.
To understand quantum machine learning models, we first need to understand quantum data and quantum classical hybrid models. Quantum data is data generated by NISQ as quantum data overlapping or intertwined with each other. Quantum data contains noise, but classical form information can be extracted as much as possible by applying quantum machine learning. TensorFlow Quantum provides the basis for generalizing quantum data in a library focused on properties.
NISQ can be used with classical processors to perform efficient calculations even when there is a lot of noise, but it is the quantum classical hybrid model that enables such calculations. TensorFlow Quantum realizes an efficient quantum classical hybrid model by mounting the NISQ compiler and scheduler included in Circue in addition to the general structures used in quantum computing such as quantum torsion, quantum gates, and quantum circuits.
When TensorFlow Quantum processes quantum data, data is evaluated first by a quantum model. Here, the classical form information hidden in the superposition of quantum data is extracted. Classical form of quantum states An arbitrary variable is a sampled or averaged neural network that processes the data by classical computing and then updates the variable to optimize the objective function.
The feature of TensorFlow Quantum is that it can simulate a relatively large quantum circuit on a multi-core processor through parallel computation in multiple quantum circuits. In order to realize the latter, it is said that it employs qsim, a high-performance quantum circuit simulator optimized for Intel multi-core processors. In this way, it is said that if the vCPU was set to 80 cores on the Google Cloud Platform N2 node, it was possible to perform 1 million simulations in 2020 quantum circuits in 60 minutes.
TensorFlow Quantum is using the existing quantum circuit simulator at the launch stage, but in the future, it is said that it will also support real quantum processors such as the quantum processor Sycamore developed by Google. TensorFlow Quantum is open source software and is also available on GitHub. Related information can be found here .