US Patent No. 11,030,548

PROBABILISTIC INFERENCE IN MACHINE LEARNING USING A QUANTUM ORACLE


Patent No. 11,030,548
Issue Date June 08, 2021
Title Probabilistic Inference In Machine Learning Using A Quantum Oracle
Inventorship Nan Ding, Los Angeles, CA (US)
Masoud Mohseni, Redondo Beach, CA (US)
Hartmut Neven, Malibu, CA (US)
Assignee Google LLC, Mountain View, CA (US)

Claim of US Patent No. 11,030,548

1. A method performed by a quantum machine, comprising:receiving, as input at a quantum machine comprising an adiabatic quantum computing system having hardware connections comprising couplers that connect qubits included in the quantum machine, data for learning an inference in a model used in machine learning, wherein:
the model is a modified restricted Boltzmann machine that includes interactions among hidden units of the restricted Boltzmann machine, wherein the interactions are based on the hardware connections of the quantum machine;
the input maps at least some interactions of different interconnected units of the model to the hardware connections between qubits in the quantum machine; and
the input is derived using data for training the model and a state of the model, the data comprising observed data for training and validating the model; and
learning, by the quantum machine, the inference in the model; and
providing, as output from the quantum machine, data representing the learned inference.