AI Google in its official blog announced the launch of TensorFlow Quantum (TFQ), which is an open source machine learning library quantum, quantum computing and machine learning can be combined together, training quantum model. Google said that this quantum machine learning models can handle the quantum data, and can be performed on a quantum computer.
The introduction of Google blog AI, allowing researchers TFQ single quantum computing data set in the FIG., The quantum and classical model configured to control parameters tensor. TensorFlow Ops will get quantum measurement result of the classical probability event, then you can use the standard functions Keras training.
Like the classic machine learning, machine learning quantum key challenge is to "Noise data" are classified. To build such a model train and substantially following steps:
- Quantum data set ready - data loaded as quantum tensor (digital multi-dimensional array). Each quantum tensor data specifies a quantum prepared Cirq circuit, this circuit generates quantum data in real time. Tensor quantum performed to generate data sets on a quantum computer TensorFlow.
- Evaluation Quantum Neural Network Model - researchers can use Cirq quantum neural networks prototyping, then embedded computing TensorFlow FIG. Quantum model essentially quantum disentangle the data inputted, so that the hidden information related to the classic coding, so that it can be used for local measurement after treatment and classic.
- Or average sample - measured quantum states in classical information extracted from the classic form of samples of a random variable. From the distribution of values of the random variable generally depends on the quantum state itself and can be observed in the measured values.
- Assessment classic neural network model - after the extraction classical information in a format suitable for the further processing of the classic.
- Evaluates a cost function - The classic process result, evaluates a cost function.
- Gradient evaluation and update parameters - direction evaluates a cost function, the cost reduction can be expected along the free parameters update pipeline.
A key feature of TensorFlow Quantum is able to simultaneously have the training and the ability to perform many quantum circuit. Currently, TensorFlow Quantum main circuit for performing quantum quantum classic circuit simulator. Google the hope that the future can be performed TFQ quantum circuits on the actual support Cirq quantum processor.
More details on TensorFlow Quantum, you can view the Google AI blog .