Information processing program, information processing method, and information processing device

Abstract

To make it easier to train variational quantum circuits used in the backdiffusion process. [Solution] The information processing device 100 generates a first data group 13(a-1) before transformation in the a-stage diffusion stage. The information processing device 100 acquires a third data group 15b generated by the second quantum circuit in accordance with the second data group 15(b-1) before transformation in the b-stage despreading stage. The information processing device 100 trains the first quantum circuit so that it identifies the first data group 13(a-1) as true and the third data group 15b as false. The information processing device 100 trains the second quantum circuit so that the trained first quantum circuit identifies the third data group 15b as true. The information processing device 100 sets the trained second quantum circuit as a variational quantum circuit 16b that represents the effect of the transformation in the b-stage despreading stage.

Claims

1. To apply this to a diffusion process that progressively transforms quantum data representing quantum states to randomize them, we obtain a first set of data representing quantum states that each follow a target distribution, To apply this to a dediffusion process that stepwise transforms quantum data representing quantum states to conform to a target distribution, we obtain a second set of data representing random quantum states, A first quantum circuit is a discriminator that identifies the truth value of quantum data representing a quantum state, and a second quantum circuit is a generator that generates false quantum data representing a quantum state. The first quantum circuit is trained so that it identifies each first data of the first data group before transformation in any diffusion stage of the diffusion process as true, and the third data generated by the second quantum circuit according to each second data of the second data group before transformation in any de-diffusion stage of the de-diffusion process corresponding to any diffusion stage as false. The second quantum circuit is trained so that the first quantum circuit, after training, identifies the third data as true. The second quantum circuit, after training, is set to a variational quantum circuit that represents the action of any of the inverse diffusion steps. An information processing program characterized by having a computer perform the processing.

2. Starting from the preceding stage of the reverse diffusion process, a combination of each reverse diffusion step and a corresponding diffusion step is selected, starting from the succeeding stage of the diffusion process. The computer is made to perform the process, The training process described above is: The information processing program according to claim 1, characterized in that each time a combination is selected, the first quantum circuit is trained to identify each first data of the first data group before transformation in the diffusion stage of the selected combination as true, the first quantum circuit is trained to identify the third data generated by the second quantum circuit in accordance with each second data of the second data group before transformation in the dediffusion stage of the selected combination as false, and the second quantum circuit is trained to identify the third data as true after training.

3. The training process described above is: An information processing program according to claim 1 or 2, characterized in that each time a combination is selected, the first quantum circuit is trained to identify each first data of the first data group before transformation in the diffusion stage of the selected combination as true, the first quantum circuit is trained to identify the third data generated by the second quantum circuit in accordance with each second data of the second data group before transformation in the dediffusion stage of the selected combination as false, and the second quantum circuit is trained to identify the third data as true after training. This series of operations is repeated until a predetermined condition is met.

4. A first cost function is set whose value decreases when the first quantum circuit identifies each of the first data in the first data group before transformation as true in the diffusion stage of the selected combination, and when the first quantum circuit identifies the third data generated by the second quantum circuit in accordance with each of the second data in the second data group before transformation as false in the dediffusion stage of the selected combination, and a second cost function is set whose value decreases when the first quantum circuit identifies the third data as true. The computer is made to perform the process, The training process described above is: The information processing program according to claim 3, characterized in that each time the above combination is selected, the parameters of the first quantum circuit are updated to train the first quantum circuit so as to minimize the set value of the first cost function, and the parameters of the second quantum circuit are updated to train the second quantum circuit so as to minimize the set value of the second cost function, and this series of operations is repeatedly performed until the predetermined conditions are met.

5. To apply this to a diffusion process that progressively transforms quantum data representing quantum states to randomize them, we obtain a first set of data representing quantum states that each follow a target distribution, To apply this to a dediffusion process that stepwise transforms quantum data representing quantum states to conform to a target distribution, we obtain a second set of data representing random quantum states, A first quantum circuit is a discriminator that identifies the truth value of quantum data representing a quantum state, and a second quantum circuit is a generator that generates false quantum data representing a quantum state. The first quantum circuit is trained so that it identifies each first data of the first data group before transformation in any diffusion stage of the diffusion process as true, and the third data generated by the second quantum circuit according to each second data of the second data group before transformation in any de-diffusion stage of the de-diffusion process corresponding to any diffusion stage as false. The second quantum circuit is trained so that the first quantum circuit, after training, identifies the third data as true. The second quantum circuit, after training, is set to a variational quantum circuit that represents the action of any of the inverse diffusion steps. An information processing method characterized in that the processing is performed by a computer.

6. To apply this to a diffusion process that progressively transforms quantum data representing quantum states to randomize them, we obtain a first set of data representing quantum states that each follow a target distribution, To apply this to a dediffusion process that stepwise transforms quantum data representing quantum states to conform to a target distribution, we obtain a second set of data representing random quantum states, A first quantum circuit is a discriminator that identifies the truth value of quantum data representing a quantum state, and a second quantum circuit is a generator that generates false quantum data representing a quantum state. The first quantum circuit is trained so that it identifies each first data of the first data group before transformation in any diffusion stage of the diffusion process as true, and the third data generated by the second quantum circuit according to each second data of the second data group before transformation in any de-diffusion stage of the de-diffusion process corresponding to any diffusion stage as false. The second quantum circuit is trained so that the first quantum circuit, after training, identifies the third data as true. The second quantum circuit, after training, is set to a variational quantum circuit that represents the action of any of the inverse diffusion steps. An information processing device characterized by having a control unit.

Images