Quantum Simulator Processor Tensor Network Contraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional quantum simulators face challenges in efficiently simulating and verifying large quantum systems, particularly in finding large sets of random amplitudes or batches of amplitudes, which is required for tasks like verifying quantum supremacy experiments, due to high computational complexity and memory requirements.
Innovation Solution
A processor configured to perform a local search algorithm to determine optimal contraction expressions for tensor networks, minimizing memory usage, computational complexity, and read-write operations, allowing for efficient contraction of tensor networks into a contracted tensor network, thereby reducing the complexity of simulating and verifying quantum circuits.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If conventional quantum simulators are used to find large sets of random amplitudes or batches of amplitudes, then the verification task can be performed, but the computational complexity and memory requirements become excessively high
Solution Approach 1:
The patent segments the tensor network contraction problem by introducing a contraction graph that divides the computation into independent sub-tensors. Each sub-tensor can be contracted separately using local search algorithms, reducing the overall computational complexity from evaluating all amplitudes simultaneously to evaluating them in parallel through segmented contractions.
Solution Approach 2:
The patent employs dynamic optimization through local search algorithms that adaptively determine contraction expressions based on the specific tensor network structure. The algorithm dynamically adjusts the contraction order and grouping to minimize computational complexity and memory requirements while maintaining the ability to find the required amplitudes.
2Quantity of substance
If conventional quantum simulators are used to find large sets of random amplitudes or batches of amplitudes, then the verification task can be performed, but the memory requirements become excessively high
Solution Approach 1:
The patent segments the tensor network into sub-tensors connected by contraction graphs, allowing memory to be allocated only for the contraction intermediates rather than storing all amplitudes simultaneously. This segmentation reduces peak memory requirements while still enabling the computation of large sets of amplitudes through systematic contraction.
Solution Approach 2:
The contraction graph acts as an intermediary structure that mediates between the input tensors and the final amplitude results. It provides a structured way to combine sub-tensor results without requiring direct storage of all intermediate amplitude values, thus reducing memory footprint while maintaining computational capability.
3Productivity
If tensor network contraction is performed without optimization, then the simulation can be completed, but the computational complexity and read-write operations increase
Solution Approach 1:
The patent uses local search algorithms to dynamically optimize the contraction expression for each tensor network. The algorithm explores different contraction orders and groupings to find the path of least computational complexity, adapting to the specific structure of each tensor network rather than using a fixed approach.
Solution Approach 2:
The patent optimizes contraction by changing parameters such as the contraction order, grouping of tensors, and selection of contraction paths. These parameter changes are made through local search algorithms that evaluate different configurations to minimize computational complexity and reduce the number of read-write operations required.
Data Source
AI summary
The present disclosure relates to the field of quantum computing, and in particular to simulating quantum circuits with a quantum simulator. The disclosure presents a processor for a quantum simulator. The processor is configured to perform a local search algorithm to determine a plurality of contraction expressions suitable to contract a respective tensor network into a determined contracted tensor network. The processor is further configured to determine, for each contraction expression, a contraction cost for contracting the respective tensor network based on a cost function, and to select the contraction expression with the lowest contraction cost to contract each tensor network into the determined contracted tensor network. The cost function is based on three parameters, which respectively indicate a required memory amount, a computational complexity, and a number of read-write operations required for contracting the respective tensor network into the determined contracted tensor network.