A quantum computer program may receive a mean, a standard deviation, and a discretization for a target normal distribution; determine a number of iterations t based on the mean and standard deviation; identify a value n from the discretization and value X.sub.0 from the mean; initialize n qubits in register q and an ancilla; apply a quantum Fourier transform and a +X.sub.0 gate to q; apply a Y-Rotation with angle π/2 to the ancilla; apply a controlled +1 gate to q controlled by the ancilla; apply a Hadamard gate to the ancilla; measure the ancilla; set a value C to a value of the measured ancilla; repeat from the Y-Rotation t times; apply an inverse quantum Fourier transform to q; and output amplitudes of basis states of q as the normal distribution. The program may use qubit-scaling to reduce the number of iterations needed.

Embodiments described herein ensure differential privacy when transmitting data to a server that estimates a frequency of such data amongst a set of client devices. The differential privacy mechanism may provide a predictable degree of variance for frequency estimations of data. The system may use a multibit histogram model or Hadamard multibit model for the differential privacy mechanism, both of which provide a predictable degree of accuracy of frequency estimations while still providing mathematically provable levels of privacy.

Disclosed by the present disclosure is a refrigerant leak detection method and device for an air conditioner. The method includes: acquiring current operating parameters of an air conditioner and environment information of a surrounding environment of the air conditioner; inputting the current operating parameters and the environment information into a trained neural network model to obtain an amount of remaining refrigerant outputted by the neural network model; and determining, according to the amount of the remaining refrigerant, whether there is a refrigerant leak in the air conditioner. The present disclosure improves accuracy of detecting a refrigerant leak in the air conditioner by means of an artificial neural network algorithm.

This invention relates to a method of classification of images among different classes comprising: performing a dimensionality reduction step for said different classes on a training set of images whose classes are known, and then classifying one or more unknown images among said different classes with reduced dimensionality, said dimensionality reduction step being performed on said training set of images by machine learning including processing, for at least a first matrix and for at least a second matrix, a parameter representative of a product of two first and second matrices to assess to which given classes several first given images respectively belong: first matrix representing the concatenation, for said several first given images, of the values of the pixels of each said first given image, second matrix representing the concatenation, for said several first given images, of the values of differences between the pixels of each said first given image and the pixels of a second given image different from said first given image but known to belong to same class as said first given image, wherein: a quantum singular value estimation is performed on first matrix, a quantum singular value estimation is performed on second matrix, both quantum singular value estimation of first matrix and quantum singular value estimation of second matrix are combined together, via quantum calculation, so as to get at a quantum singular value estimation of said product of both first and second matrices, said quantum singular value estimation of said product of both first and second matrices being said parameter representative of said product of two first and second matrices processed to assess to which given classes said several first given images respectively belong.

The embodiment applies a two-level coding/decoding technology using a spread spectrum code and an orthogonal code, whereby touch sensing can be robust against noise, a touch sensing time can be reduced, and touch sensitivity can be enhanced.

A method of providing transformed target points for integrating a component into an assembly includes collecting a set of component target points, collecting a set of assembly target points, identifying target points common to the set of component target points and the set of assembly target points; performing a specified number of Monte Carlo transformations of selected ones of the common target points to yield a set of transformed target points and vectors and an associated uncertainty value for each transformed target point and vector, and using certain ones of the transformed target points for integrating the component into the assembly based on the associated uncertainty value for each of the transformed target points.

The invention relates to methods, devices, systems and computer program products where input image data is received for processing in a computer system, a first transform is carried out to the input image data, the transform being done in a first direction with respect to the input image data, to obtain first transformed image data, a transposing operation is carried out to the first transformed image data to obtain first transposed image data, and a second transform is carried out to the first transposed image data, the second transform being done in a second direction with respect to the input image data, to obtain second transformed image data. The first transform and the second transform may be carried out in a processor in a parallel manner, and the transposing operation may be carried out at least partially using registers of the processor.

A method for adaptive stochastic learning state compression for federated learning in infrastructure domains. Specifically, the disclosed method introduces an adaptive data compressor directed to reducing the amount of information exchanged between nodes participating in the optimization of a shared machine learning model through federated learning. The adaptive data compressor may employ stochastic k-level quantization, and may include functionality to handle exceptions stemming from the detection of unbalanced and/or irregularly sized data.

The embodiment applies a two-level coding/decoding technology using a spread spectrum code and an orthogonal code, whereby touch sensing can be robust against noise, a touch sensing time can be reduced, and touch sensitivity can be enhanced.

A computer-implemented method is presented for performing a computation with a neural network. The method includes: receiving a first input patch of data; applying a Walsh-Hadamard transform to the input patch to yield a transformed input patch in a transformed domain; computing an element-wise product of the transformed input patch and a kernel of the neural network; applying an inverse Walsh-Hadamard transform to the element-wise product to yield an intermediate matrix; and creating a first output patch from the intermediate matrix, where the size of the first output patch is smaller than the intermediate matrix.