Multiscale contiguous block pixel entangler for image recognition on hybrid quantum-classical computing system
A method of performing implementing a quantum convolutional neural network (QCNN) in a hybrid quantum-classical computing system includes performing a data load operation, a set of a convolutional layer operation and a pooling operation, a measurement operation. The data load operation includes encoding pixel data of an input image onto a quantum processor using column qubits and row qubits. The convolutional layer operation includes a contiguous block pixel entangler that entangles a column qubit and a row qubit, depending on a pattern of a feature to detect in the input image. The pooling layer operation includes applying a series of one-qubit operations to the column qubits and the row qubits. The measurement operation includes measuring a state of an output qubit among the column qubits and the row qubits.
1 . A method of implementing a quantum convolutional neural network (QCNN) in a hybrid quantum-classical computing system comprising a classical computer, a quantum processor, and a system controller, the method comprising:
performing a data load operation comprising:
receiving, by the classical computer, an input image having a two dimensional array of pixels arranged in columns and rows; and
encoding, by the system controller, pixel data of the input image onto the quantum processor using a plurality of column qubits and a plurality of row qubits, wherein the plurality of column qubits represents pixels along the column direction and the plurality of row qubits represents pixels along the row direction;
performing a set of a convolutional layer operation and a pooling layer operation,
the convolutional layer operation comprising applying, by the system controller, a contiguous block pixel entangler that entangles a column qubit of the plurality of column qubits and a row qubit of the plurality of row qubits, depending on a pattern of a feature to detect in the input image; and
the pooling layer operation comprising applying, by the system controller, a series of one-qubit operations to the plurality of column qubits and the plurality of row qubits;
performing a measurement operation comprising measuring, by the system controller, a state of an output qubit among the plurality of column qubits and the plurality of row qubits; and
outputting, by the classical computer, qubit state information relating to the measured state of the output qubit, wherein the qubit state information includes information relating to whether or not the feature is contained in the input image.
2 . The method of claim 1 , wherein the contiguous block pixel entangler entangles pixels arranged diagonally in a direction between the column direction and the row direction in the input image.
3 . The method of claim 1 , wherein the contiguous block pixel entangler entangles pixels arranged horizontally in the column direction in the input image.
4 . The method of claim 1 , wherein the contiguous block pixel entangler entangles pixels arranged vertically in the row direction in the input image.
5 . The method of claim 1 , wherein the contiguous block pixel entangler applied to a first column qubit of the plurality of column qubits and a first row qubit of the plurality of row qubits entangles pixels in a block of 2×2 contiguous pixels in the input image.
6 . The method of claim 5 , wherein the contiguous block pixel entangler applied to the first column qubit of the plurality of column qubits and a second row qubit of the plurality of row qubits entangles pixels in a block of 2×4 contiguous pixels in the input image.
7 . The method of claim 6 , wherein the contiguous block pixel entangler applied to a second column qubit of the plurality of column qubits and the first row qubit of the plurality of row qubits entangles pixels in a block of 4×2 contiguous pixels in the input image.
8 . The method of claim 1 , wherein the set of the convolutional layer operation and the pooling layer operation is repeated until the information about whether or not the feature is contained in the input image is contained in the output qubit.
9 . A hybrid quantum-classical computing system, comprising:
a quantum processor comprising a plurality of trapped ions, each of the trapped ions having two hyperfine states defining a qubit;
a system controller configured to control one or more lasers configured to emit a laser beam, which is provided to trapped ions in the quantum processor; and
a classical computer configured to:
receive an input image having a two dimensional array of pixels arranged in columns and rows;
control the system controller to encode pixel data of the input image onto the quantum processor using a plurality of column qubits and a plurality of row qubits, wherein the plurality of column qubits represents pixels along the column direction and the plurality of row qubits represents pixels along the row direction;
control the system controller to perform a set of a convolutional layer operation and a pooling layer operation,
the convolutional layer operation comprising applying a contiguous block pixel entangler that entangles a column qubit of the plurality of column qubits and a row qubit of the plurality of row qubits, depending on a pattern of a feature to detect in the input image; and
the pooling layer operation comprising applying a series of one-qubit operations to the plurality of column qubits and the plurality of row qubits;
control the system controller to perform a measurement operation comprising measuring a state of an output qubit among the plurality of column qubits and the plurality of row qubits; and
output qubit state information relating to the measured state of the output qubit, wherein the qubit state information includes information relating to whether or not the feature is contained in the input image.
10 . The hybrid quantum-classical computing system of claim 9 , wherein
each of the trapped ions is 171 Yb + having the 2 S 1/2 hyperfine states.
11 . The hybrid quantum-classical computing system of claim 9 , wherein
each of the trapped ions is one selected from Be + , Ca + , Sr + , Mg+, Ba + , Zn + , Hg + , Cd + .
12 . The hybrid quantum-classical computing system of claim 9 , wherein the contiguous block pixel entangler entangles pixels arranged diagonally in a direction between the column direction and the row direction in the input image.
13 . The hybrid quantum-classical computing system of claim 9 , wherein the contiguous block pixel entangler entangles pixels arranged horizontally in the column direction in the input image.
14 . The hybrid quantum-classical computing system of claim 9 , wherein the contiguous block pixel entangler entangles pixels arranged vertically in the row direction in the input image.
15 . The hybrid quantum-classical computing system of claim 9 , wherein the contiguous block pixel entangler applied to a first column qubit of the plurality of column qubits and a first row qubit of the plurality of row qubits entangles pixels in a block of 2×2 contiguous pixels in the input image.
16 . The hybrid quantum-classical computing system of claim 15 , wherein the contiguous block pixel entangler applied to the first column qubit of the plurality of column qubits and a second row qubit of the plurality of row qubits entangles pixels in a block of 2×4 contiguous pixels in the input image.
17 . The hybrid quantum-classical computing system of claim 16 , wherein the contiguous block pixel entangler applied to a second column qubit of the plurality of column qubits and the first row qubit of the plurality of row qubits entangles pixels in a block of 4×2 contiguous pixels in the input image.
18 . A hybrid quantum-classical computing system comprising non-volatile memory having a number of instructions stored therein which, when executed by one or more processors, causes the hybrid quantum-classical computing system to perform operations comprising:
performing a data load operation comprising:
receiving, by a classical computer, an input image having a two dimensional array of pixels arranged in columns and rows; and
encoding, by a system controller, pixel data of the input image onto a quantum processor using a plurality of column qubits and a plurality of row qubits, wherein the plurality of column qubits represents pixels along the column direction and the plurality of row qubits represents pixels along the row direction;
performing a set of a convolutional layer operation and a pooling layer operation,
the convolutional layer operation comprising applying, by the system controller, a contiguous block pixel entangler that entangles a column qubit of the plurality of column qubits and a row qubit of the plurality of row qubits, depending on a pattern of a feature to detect in the input image; and
the pooling layer operation comprising applying, by the system controller, a series of one-qubit operations to the plurality of column qubits and the plurality of row qubits;
performing a measurement operation comprising measuring, by the system controller, a state of an output qubit among the plurality of column qubits and the plurality of row qubits; and
outputting, by the classical computer, qubit state information relating to the measured state of the output qubit, wherein the qubit state information includes information relating to whether or not the feature is contained in the input image.
19 . The hybrid quantum-classical computing system of claim 18 , wherein the contiguous block pixel entangler entangles pixels arranged diagonally in a direction between the column direction and the row direction in the input image, pixels arranged horizontally in the column direction in the input image, or pixels arranged vertically in the row direction in the input image.
20 . The hybrid quantum-classical computing system of claim 18 , wherein
the contiguous block pixel entangler applied to a first column qubit of the plurality of column qubits and a first row qubit of the plurality of row qubits entangles pixels in a block of 2×2 contiguous pixels in the input image,
the contiguous block pixel entangler applied to the first column qubit of the plurality of column qubits and a second row qubit of the plurality of row qubits entangles pixels in a block of 2×4 contiguous pixels in the input image, and
the contiguous block pixel entangler applied to a second column qubit of the plurality of column qubits and the first row qubit of the plurality of row qubits entangles pixels in a block of 4×2 contiguous pixels in the input image.