6 results about "Linear solver" patented technology

The embodiment of the invention discloses a parameter training method of a variable component sub-linear solver and a related device. The method comprises the following steps: calculating the gradient of a quantum circuit in the variable component sub-linear solver under a variable parameter at the current moment; if the norm of the gradient at the current moment is greater than or equal to a preset closing precision standard, determining a Hessian approximation matrix at the current moment according to the variable parameter increment and the gradient increment of the current moment relative to the previous moment; determining a variable parameter of the next moment according to the variable parameter of the current moment, the gradient and the Hessian approximation matrix; and taking the variable parameter of the next moment as the variable parameter of the current moment, and executing the step of calculating the gradient of the quantum circuit under the variable parameter of the current moment. By adopting the embodiment of the invention, the parameter training efficiency of the variable component sub-linear solver can be improved while computer hardware resources are saved.

This invention discloses a method, apparatus, and medium for solving variable quantum linear problems in subspaces. The method includes: determining the system of linear equations to be solved, constructing a variable quantum circuit and a Krylov subspace, and using the generalized minimum residual method and the variable quantum circuit to calculate an approximate solution of the system of linear equations to be solved in the subspace. This addresses the shortcomings of existing technologies and can reduce the time complexity and computational load of solving linear problems.

This application provides a nonlinear finite element accelerated solution system and method. The system includes: a physical information neural network model construction module configured to: acquire configuration points in the solution domain of the nonlinear partial differential equation to be solved; input the configuration points into a fully connected neural network to obtain a predicted solution; and obtain a physical information neural network model based on the predicted solution; an input module configured to: construct a finite element model of the nonlinear partial differential equation to be solved; transform the nonlinear partial differential equation to be solved into a system of nonlinear algebraic equations; and generate a predicted value vector based on the finite element model; and a solution module configured to: solve the system of nonlinear algebraic equations using a nonlinear solver to obtain the target solution. This addresses the problem that in current finite element methods for parameterized research scenarios, each time a new parameter value is input, a complete and computationally expensive nonlinear finite element solution process needs to be started independently, leading to repetitive calculations and low efficiency.

The embodiment of the invention discloses a Helmholtz equation quantum solving method and an electromagnetic wave characteristic determination method and device, and the method comprises the steps: carrying out the series expansion of a solution and a source item of a Helmholtz equation according to a selected primary function, and discretizing a computational domain of the Helmholtz equation into grid points; substituting the solution and the series form of the source item into a margin function corresponding to the Helmholtz equation, and applying a weighted margin method at each network point to obtain a large linear equation set of which the dimension is consistent with the number of grid points; solving a linear equation set through a quantum linear solver to obtain a solution coefficient of the primary function, wherein the linear equation set is obtained by applying a weighted margin method to all grid points; and substituting the solution coefficient of the primary function into the series form of the solution to obtain the solution. By adopting the embodiment of the invention, the hardware resources of the computer can be saved, and the solving speed of the Helmholtz equation can be improved.

The invention relates to the technical field of computer-aided engineering and building information models, and discloses a fire-fighting installation BIM model and three-dimensional scanning intelligent acceptance and verification method, which comprises the following steps: constructing a discretization rigidity constraint model reflecting pipe network mechanical response, converting the space position difference between the actually measured pipe network center track and the design model into a displacement load vector; resolving the reverse elastic equilibrium equation by using a finite element linear solver to obtain internal force distribution and node torque response in a pipe network matching state; according to the method, a reverse mapping mechanism of geometric deviation and mechanical load is established, traditional geometric comparison is upgraded to assembly stress calculation, flexible adaptation and forced assembly are distinguished, and engineering potential safety hazards caused by hidden stress concentration are avoided.

The invention discloses a method and a device for adiabatic quantum calculation of a linear equation set based on block coding. The method comprises the following steps: acquiring the linear equation set Ax = b; constructing a time-dependent Hamiltonian H (f (s)) of adiabatic evolution, wherein the eigenstate of the initial Hamiltonian H (0) is expressed as bgt; the eigenstate of the target Hamiltonian H (1) is expressed as A-1bgt; ; a discrete adiabatic evolution quantum circuit is constructed, the discrete adiabatic evolution quantum circuit has an evolution operator UH (f (s)) corresponding to the time-dependent Hamiltonian H (f (s)), the evolution operator UH (f (s)) continuously acts on the discrete adiabatic evolution quantum circuit, and the initial quantum state corresponds to the eigenstate bgt of the adiabatic evolution initial Hamiltonian H (0); the final quantum state corresponds to the eigenstate A-1bgt of the adiabatic evolution target Hamiltonian H (1); . Compared with the prior art, the invention provides a specific implementation method for a quantum circuit of Hamiltonian gauge block coding, construction of a discrete adiabatic quantum linear solver can be realized based on the block coding, and solving of a linear equation set is completed.