US 2026/0187185 A1 · TENSOR NETWORK ANNEALING METHOD AND SYSTEM FOR OPTIMIZATION OF MACHINE LEARNING TASKS

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. The system includes a computer program product and a computer-readable data carrier for executing this method, providing a robust framework for annealing tasks and optimization using memory and processor configurations designed for matrix product state representation and compression.

Claims15

Dependent claims

A computer-implemented method for computational optimization tasks in machine learning, and data science, comprising:

storing at least one vector within a memory medium, configured to optimize storing data and to configure parameters of a processor and defining vector-scalar multiplication followed by a vector addition parameter on the multi-variate binary function;
and
configuring the processor to run as an annealer and perform:
obtaining a multi-variate binary function which is a data that is stored within the memory medium;
creating data in a matrix product state (MPS) format from the multi-variate binary function by generating a representation through matrix multiplications selected from matrices corresponding to binary variables with an MPS generator;
constructing the matrix product state with a bond dimension with the MPS constructor;
decomposing at least one vector in the matrix product state structure of the data with the MPS decomposition module to achieve optimal decomposition;
applying maximal bond dimension to the tensor train of the matrix product state according to the number of variables;
and
determining and storing a final state of the tensor train within the memory medium.

7 more elements

Dependencies are inferred from the claim text.

Tensor network annealing method and system for optimization of machine learning tasks

Life of this patent

Tensor network annealing method and system for optimization of machine learning tasks

Life of this patent

Abstract

Claims15