# US 2025/0209374 A1 · System and methods for implementing symmetric tensor networks for quantum machine learning and related methods

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2025/0209374 A1
- Document ID: US20250209374A1
- Status: Published application
- Kind: A1
- Application number: 18/398,879
- Filed: 2023-12-28
- Published: 2025-06-26
- Earliest claimed: 2023-12-20
- Assignees: Multiverse Computing S.L.
- Applicants: Multiverse Computing S.L.
- Inventors: Roman ORÚS; Saeed JAHROMI; Sukhbinder SINGH
- Web page: https://patentgrep.com/patent/US20250209374A1

## CPC

- G06N 3/063: using electronic means
- G06N 20/00: Machine learning
- G06N 3/09: Supervised learning
- G06N 10/00: Quantum computing, i.e. information processing based on quantum-mechanical phenomena
- G06N 10/60: Quantum algorithms, e.g. based on quantum optimisation, quantum Fourier or Hadamard transforms
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A computational framework and a method for implementing symmetric tensor networks for quantum machine learning using symmetric deep learning involve the execution of techniques to deploy a computational framework, creation of mathematical structures for machine learning, and utilization of deep learning for faster convergence and training. The system includes hardware components to implement the model, an optimization technique to adjust elements, and an inference module for predictions over new datapoints. The mathematical structures used are symmetric tensor networks, which are built using the symmetries of the dataset and replace the weight matrices in the deep learning module.

## Claims

### Claim 1

A computational framework, comprising:
execution of techniques or procedures to deploy a computational framework;
creation of mathematical structures for utilization of a type of machine learning;
utilization of a type of deep learning for faster convergence and training, better precision;
employment of hardware components to implement the model;
extraction of data from a collection of data;
arrangement of computational units in layers;
substitution of numerical arrays in deep learning module by mathematical structures;
usage of specific type of mathematical structures;
application of optimization technique to adjust adjustable elements;
reduction of performance measure for a subset of dataset; and
generation of output of the model for unseen data.

### Claim 2 (depends on 1)

The computational framework of claim 1, wherein the mathematical structures are symmetric tensor networks.

### Claim 3 (depends on 2)

The computational framework of claim 2, wherein the symmetric tensor networks are built using the symmetries of the dataset.

### Claim 4 (depends on 3)

The computational framework of claim 3, wherein the weight matrices in the deep learning module are replaced by the symmetric tensor networks.

### Claim 5 (depends on 4)

The computational framework of claim 4, wherein the system includes a classical optimization algorithm that fine-tunes the parameters of the symmetric tensor deep learning network to minimize a cost function for a training set.

### Claim 6 (depends on 5)

The computational framework of claim 5, wherein the system includes an inference module that makes predictions over a new set of datapoints.

### Claim 7 (depends on 6)

The computational framework of claim 6, wherein the model can be implemented on deep learning chips.

### Claim 8

A method for implementing symmetric tensor networks for quantum machine learning using symmetric deep learning, the method includes the following steps:
execution of techniques or procedures to deploy a computational framework;
creation of mathematical structures for utilization of a type of machine learning;
utilization of a type of deep learning for faster convergence and training, better precision;
employment of hardware components to implement the model;
extraction of data from a collection of data;
arrangement of computational units in layers;
substitution of numerical arrays in deep learning module by mathematical structures;
usage of specific type of mathematical structures;
application of optimization technique to adjust adjustable elements;
reduction of performance measure for a subset of dataset; and
generation of output of the model for unseen data.

### Claim 9 (depends on 8)

The method of claim 8, wherein the mathematical structures are symmetric tensor networks.

### Claim 10 (depends on 9)

The method of claim 9, wherein the symmetric tensor networks are built using the symmetries of the dataset.

### Claim 11 (depends on 10)

The method of claim 10, wherein the weight matrices in the deep learning module are replaced by the symmetric tensor networks.

### Claim 12 (depends on 11)

The method of claim 11, wherein the system includes a classical optimization algorithm that fine-tunes the parameters of the symmetric tensor deep learning network to minimize a cost function for a training set.

### Claim 13 (depends on 12)

The method of claim 12, wherein the system includes an inference module that makes predictions over a new set of datapoints.

### Claim 14 (depends on 13)

The method of claim 13, wherein the model can be implemented on deep learning chips.

### Claim 15 (depends on 14)

The method of claim 14, wherein the symmetric tensor networks are symmetric matrix product operators.

### Claim 16 (depends on 15)

The method of claim 15, wherein the classical optimization algorithm uses backpropagation or similar.

### Claim 17 (depends on 16)

The method of claim 16, wherein the parameters are fine-tuned to minimize a cost function for a training set.

### Claim 18 (depends on 17)

The method of claim 17, wherein the inference module makes predictions over a new set of datapoints.

### Claim 19 (depends on 18)

The method of claim 18, wherein the predictions are made over a new set of datapoints.

### Claim 20 (depends on 19)

The method of claim 19, wherein the new set of datapoints are unseen data.
