Tensor Network Parallel Simulation for Large-Scale Dynamical Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for simulating large-scale dynamical systems are inefficient due to scalability issues in existing commercial software, which often require system size reduction and neglect high-order system modes, limiting their accuracy and reliability in industrial applications.
Innovation Solution
A parallel algorithm based on tensor network methods that numerically solve large systems of ordinary linear differential equations using Suzuki Trotter decomposition, allowing for distributed computing and reducing data transfer among cores, thereby maintaining high accuracy without system size reduction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing commercial software is used for simulating large-scale dynamical systems, then the simulation can be performed with standard tools, but the scalability is poor and system size reduction is required which limits accuracy
Solution Approach 1:
The patent divides the large-scale dynamical system into multiple smaller subsystems that can be simulated independently using tensor network methods. This segmentation allows the simulation to maintain high accuracy without requiring reduction of the overall system size, as each subsystem can be processed separately while preserving the full system's dynamics through the tensor network framework
Solution Approach 2:
The patent transforms the simulation approach by changing the mathematical parameters and representation of the system. By using tensor network decompositions and Suzuki Trotter decomposition, the system transforms the computational problem into a form that scales efficiently, changing how the system equations are represented and solved rather than reducing the system itself
2Productivity
If system size reduction is applied to make simulation feasible, then computational resources are reduced, but high-order system modes are neglected reducing accuracy
Solution Approach 1:
The patent introduces a new dimensional approach by representing the system in tensor network space rather than traditional state-space representation. This dimensional transformation allows the simulation to handle high-order modes efficiently by distributing them across the tensor network structure, maintaining accuracy while improving computational efficiency through the inherent parallelism of the tensor representation
3Loss of energy
If parallel computing is implemented using tensor network methods, then data transfer among cores is reduced, but the implementation complexity increases
Solution Approach 1:
The patent segments the computational workload across multiple processor cores in a parallel architecture, with each core handling specific tensor network operations. This segmentation reduces data transfer overhead by localizing computations and minimizing inter-core communication, as the tensor network structure naturally divides the problem into independent computational units that can be processed in parallel
Solution Approach 2:
The tensor network structure itself acts as an intermediary that facilitates parallel computation. By representing the system dynamics through tensor contractions and transformations, the patent creates an intermediate mathematical framework that enables efficient parallel processing while reducing the need for extensive data exchange between cores compared to traditional methods
Data Source
AI summary
A system includes a memory storing computer-readable instructions and at least one processor to execute the instructions to perform at least one tensor network method to numerically solve at least one differential equation.