Abstract: An apparatus or system comprising at least one classical processor and/or at least one quantum processor configured to calculate a first cost of a predetermined cost function associated with both a predetermined optimization problem and a set of variables as parameters, and calculate the cost of the cost function a plurality of times running an optimization by applying gradient descent to a converted set of variables a plurality of times considering the most recent calculated cost of the cost function in each conversion of the set of variables.

Abstract: A method comprising: determining, in an AI system comprising a deep learning model with a plurality of layers defining a weight matrix of the model, groups of two or more contiguous layers in the plurality of layers; for each determined group, applying SVD to weights of one layer of the group; for each layer with SVD applied thereto, truncating at least some singular values obtained from the application of the SVD based on at least one truncation criterion; for each determined group, selecting a plurality of output channels of the deep learning model for removal based on at least one weight importance criterion; and for each determined group, removing one or more input channels and/or one or more output channels based on the selected plurality of output channels such that a shared dimension size in the group is maintained.

Abstract: A computer-implemented method and an apparatus for automatic modulation recognition that enables the detection and identification of modulation schemes in received raw signals with a signal receiving unit without prior information about the raw signal detail, comprising a computing unit configured to transform received raw signals from time domain to frequency domain including the noise in the signal with segmenting the signal and computing its modulation in multiple image with the spectrogram extraction process in order to capture temporal dependencies and sequential information by treating the raw signals as images; augmentation of the data for increasing the dataset size for increasing the accuracy with the limited data; training the data for enabling the network to learn spatiotemporal relationships based on a signal-to-noise ratio level; applying the data to one algorithm of two, which are convolutional neural network and convolutional neural network long short-term memory network hybrid algorithm.

Abstract: Random number generation is crucial in applications such as cryptography, simulations, and statistical sampling. However, traditional methods often rely on algorithmic processes, which may not provide true randomness. An example solution may provide a computer-implemented method including: receiving a data set, a processing time, and a state count; executing one or more simulations of a quantum adiabatic process based on the data set, the processing time, the state count, an energy function, and one or more network structures, the one or more network structures including a representation of one or more initial simulated quantum bits; measuring one or more simulated values based on the one or more evolved simulated quantum bits at the end of each simulation; and outputting one or more output values based on the one or more simulated values, the one or more output values including one or more classical bit.

Abstract: A computer-implemented method for performing computational optimization in machine learning and data science involves a system equipped with at least one processor and memory. This method encompasses storing vectors, defining vector-scalar multiplication, and configuring the processor to operate as an annealer. Key steps include transforming a multi-variable binary function into a matrix product state (MPS) format, constructing the MPS with a specific bond dimension, and decomposing vectors for optimal efficiency. The method employs a maximal bond dimension for the tensor train, determining its final state, and utilizes tensor network compression techniques to compress the matrix product state. It addresses Quadratic Unconstrained Binary Optimization, frequently used in machine learning, to optimize binary functions.

Abstract: A computer-implemented method for recovering a plaintext is disclosed. The plaintext (300) comprises a variable string (320). The method comprises constructing (S2) a tensor network (500) representing the variable string (320), optimizing (270) the tensor network (500), generating (S3), in a hash function generator (350), a candidate hash (370) based on the output of the tensor network (330), determining (S4) an overlap (390) between the candidate hash (370) and a reference ciphertext (380); and on reaching a zero-overlap value, determining the variable string (320), otherwise optimising the tensor network (500).

Abstract: A method comprising: processing an instruction to modify a trained artificial intelligence, AI, system, the trained AI system having the form of a tensor network, the instruction comprising adding a first dataset in the AI system and/or removing a second dataset from the AI system; and running the instruction on the trained AI system. When the instruction comprises adding the first dataset, encoding data values of the first dataset in one or more tensors of the trained AI system associated with values of the first dataset. When the instruction comprises removing the second dataset, setting to zero values associated with data values or relationships of the second dataset in one or more tensors of the trained AI system associated with values of the second dataset, or removing the one or more tensors associated with values of the second dataset from the AI system.

Abstract: A computer-implemented method for training a machine learning model using a training dataset having a set of distributional parameters ? is disclosed. The method comprises selecting a subset from the training dataset, calculating a gradient of the subset by encoding a cost function representative of the gradient into a quantum circuit, amplifying the amplitude of the quantum circuit, constructing a likelihood function, and minimising the cost function in a variational quantum circuit to optimise the distributional parameters, extracting the optimised distributional parameters, and entering the optimised distributional parameters into the machine learning model.

Abstract: A device for editing large language models and multimodal AI models. The device includes a memory medium that stores the data and/or dataset of the model, a control unit that is configured to process the data within modules and according to requests and pre-defined configuration parameters from a user and/or client which communicates with the device with at least one communication unit to provide at least one output based on an input data; and runs the instruction on the trained AI model. The device includes a compression module, a lobotomization module and an inception module and at least one of the modules is configured to identify layers of the trained model with the weight matrices; and decomposing the weight matrices of the trained model into a tensor network structure. The device also includes an integration module that provides connection and data transfer between external systems and platforms and the device.

Abstract: Reliance on cloud-based decision-making presents several challenges. To address at least some of these technical challenges, an example system is provided. The system includes an edge computing device including a decision-making model. The system also includes at least one local sensor coupled to the edge computing device through a localized network, wherein the at least one local sensor is disposed in a physical environment. The system also includes at least one local device coupled to the edge computing device through the localized network, wherein the at least one local device is disposed in the physical environment. The edge computing device is configured to: receive, from the at least one local sensor, audio data associated with the physical environment; determine, responsive to the audio data, a state associated with the physical environment using the decision-making model; and determine, based on the state, an instruction for the at least one local device.

Abstract: The present invention proposes a computer implemented method and system for determining a cryptographic key. The method comprises constructing a tensor network with parameters representing a candidate cryptographic key; adjusting the parameters of the tensor network; generating a candidate key sample and obtaining a candidate ciphertext obtained with the candidate cryptographic key; calculating a cost function with respect to a target ciphertext, measuring an overlap between the target ciphertext and the candidate ciphertext, determining whether the overlap has reached a threshold value. If threshold value is not reached, repeating the method by further adjusting the parameters of the tensor network, if the threshold value is reached, determining that the candidate cryptographic key is the cryptographic key.

Abstract: Reliance on cloud-based decision-making presents several challenges. To address at least some of these technical challenges, an example system is provided. The system includes an edge computing device including a decision-making model. The system also includes at least one local sensor coupled to the edge computing device through a localized network, wherein the at least one local sensor is disposed in a physical environment. The system also includes at least one local device coupled to the edge computing device through the localized network, wherein the at least one local device is disposed in the physical environment. The edge computing device is configured to: receive, from the at least one local sensor, image data associated with the physical environment; determine, responsive to the image data, a state associated with the physical environment using the decision-making model; and determine, based on the state, an instruction for the at least one local device.

Abstract: A computer-implemented method for simulating and analyzing material deformations with principles of linear elasticity, comprising: splitting material properties of a material into pieces; decomposition of the geometry into simple domains and computing mesh-grid tensors for all dimensions for specifying a corresponding coordinate with a defining mesh-grid function; using the mesh-grid tensors for each simple domains that are defined for all available dimensions to compute components of a Jacobian, and mapping a reference element into a small element of each simple domain to compute the components all at once; processing stress and strain fields of the material, external forces applied thereto and test function equations over local elements through the Jacobian components; assembling stiffness TN-operators and mass TN-operators for each simple domain; stitching the stiffness TN-operators and the mass TN-operators together for enforcing continuity over interfaces; and applying a mass TN-operator to a predefined

Abstract: A method including the following steps: computing availability of at least one renewable electrical energy production device that delivers electricity to at least one household associated therewith to deliver electricity to an electric grid; computing an electricity usage plan for the predetermined period of time; optimizing the electricity usage plan so that: the electricity to be delivered is to be delivered to the electric grid at one or more different time periods at least when an amount of the possible electric power generation corresponding to non-renewable electrical energy generators is greater than at other time periods within the predetermined period of time; and during a predetermined duration after the time period or periods when the electricity is delivered to the electric grid, the electricity delivered by the at least one renewable electrical energy production device for consumption by the respective household is lower than the computed availability during the predetermined duration.

Abstract: The method enhances computational efficiency and performance of large language models (LLMs) through hybrid classical-quantum processing. The method involves a classical computer receiving an input for processing by a selected LLM comprising deep self-attention and multilayer perceptron layers and decomposing the LLM's weight matrices into a first Matrix Product Operator (MPO). The classical computer identifies one or more disentanglers which factorize the MPO into a non-unitary tensor network and a set of unitary subcomponents. The non-unitary tensor network enables compression by localizing quantum correlations. The unitary subcomponents correspond to first and second variational quantum circuits, configured to run sequentially on a quantum computer. The execution of the second quantum circuit is conditioned on both the measurements of the first quantum circuit and outputs from the tensor network, enabling enhanced correlation capture and efficient model representation.

Abstract: A method includes: receiving, by an apparatus or system, a first set of values for an electric grid, a second set for a first electricity supply, and a third set for a second electricity supply; receiving data from electricity controllers, wherein all the received data form a fourth set of values, wherein each electricity controller of the controllers is associated with a household of multiple households and controls electricity supplied to the electric grid, stored and requested from the electric grid at the different time periods within the PT, wherein each household includes at least one renewable electrical energy production device and the fourth set of values is indicative of electrical energy producible by each device of the at least one renewable energy production device at the different time periods within the PT; solving, by the apparatus or system based on the four sets of values, an optimization problem.

Abstract: A computer implemented method for solving a classical optimization problem of integer factorization implemented on a digital computer system is described. The method is implemented on a classical processor adapted to execute a time evolving block decimation algorithm. The method comprises in a first step an inputting a lattice basis and a target lattice vector to an input device of the classical processor followed by an implementing a lattice basis reduction algorithm on the lattice basis in an implementation module, thereby obtaining a reduced orthogonal lattice basis. The method further comprises a projecting the target lattice vector on the reduced orthogonal lattice basis followed by a building a closest vector to the target lattice vector and optimizing the closest vector using a tropical time-evolving block decimation algorithm by the classical processor and finally outputting an integer vector.

Abstract: A method for modifying a quantum neural network includes the steps of: identifying via a computer information that needs to be erased from a trained quantum neural network; localizing via the computer the identified information in the trained quantum neural network; erasing via the computer the identified information from the trained quantum neural network without erasing from the trained quantum neural network other information that needs not to be erased; and compressing via the computer the quantum neural network which results when the identified information has been erased. A system including similar components is related.

Abstract: A system configured for: breaking down a cost function into two-bit terms; applying a two-bit gate of temporal evolution and step change to a selected two-bit term, thereby generating a tensor per bit, with a connecting tensor index between the two tensors; shortening the connecting tensor index; removing at least as many connecting tensors as needed to reduce the number of connecting tensor indices down to a threshold M; and applying, shortening and removing for one, some or all other two-bit terms of the cost function at least until a solution to the cost function reaches a predetermined convergence.

Abstract: A computer-implemented method and an apparatus for automatic modulation recognition that enables the detection and identification of modulation schemes in received raw signals with a signal receiving unit without prior information about the raw signal detail, comprising a computing unit configured to transform received raw signals from time domain to frequency domain including the noise in the signal with segmenting the signal and computing its modulation in multiple image with the spectrogram extraction process in order to capture temporal dependencies and sequential information by treating the raw signals as images; augmentation of the data for increasing the dataset size for increasing the accuracy with the limited data; training the data for enabling the network to learn spatiotemporal relationships based on a signal-to-noise ratio level; applying the data to one algorithm of two, which are convolutional neural network and convolutional neural network long short-term memory network hybrid algorithm.