14 results about "Descent direction" patented technology

The invention relates to a method, a system and equipment for adjusting resonance integral in a physical multi-group constant library of a reactor core, and the method comprises the following steps: (S1) determining a descent gradient of a steepest descent method for optimizing the resonance integral for a selected sample experiment group; (S2) determining an optimized and adjusted resonance integral according to a gradient descent direction of a steepest descent method, and merging the resonance integral into a main library of target nuclides; (S3) making subgroup parameters based on the optimized and adjusted resonance integral; (S4) generating an optimized and adjusted reactor core physical multi-group constant library; and (S5) carrying out benchmark inspection on the optimized reactor core physical multi-group constant library. According to the method, the accuracy of effective multiplication factors and other integral quantities obtained through calculation of the reactor core physical multi-group constant library based on the subgroup method during reactor physical calculation can be ensured, and meanwhile, the universality of the reactor core physical multi-group constant library is also expanded.

The invention provides an inversion method and system based on a hybrid regularization model, electronic equipment and a storage medium. The method comprises the following steps: S1, inputting seismic data, and constructing a target function; s2, judging whether the number of iterations of the target function reaches the maximum value Kmax and whether the error of two times of iterations reaches a small value or not; s3, calculating a pseudo gradient # imgabs0 # of the target function according to a judgment result of the step 2; S4, calculating a descending direction of the pseudo gradient # imgabs1 # by utilizing L-BFGS loop iteration, and mapping the same quadrant for correction; and S5, according to the descending direction of the pseudo gradient # imgabs2, calculating a new inversion result through a linear search strategy. According to the method, through mixed regularization constraint inversion, the development condition of the stratum is better described, and the OWL-QN algorithm is introduced to accelerate the convergence of the algorithm and improve the inversion precision.

The invention discloses an intelligent design method and system for a machining clamp based on Zhongwishi 3D. The core innovation of the method is that a'dynamic time variant prime field 'and'reverse differential evolution' mechanism is constructed; in multi-physics field simulation of virtual manufacturing, interference potential energy of a tool time sequence envelope and a constraint unit is calculated in real time; a Jacobian matrix of a parameter space and a potential energy space is established, the steepest descent direction of the interference gradient is analyzed, and a Newton-Raphson iteration method is used for driving a tool model to conduct self-adaptive geometric deformation or pose transformation. According to the method, full-process automation from geometric analysis, rule reasoning and physical field simulation to parameter closed-loop optimization is achieved, the problems of CAD / CAM data fault and open-loop iteration in traditional tool design are solved, and mathematical-level global convergence and machining safety of technological equipment are ensured.

According to the fusion type robot path planning method based on channel guidance, an RRT-star module provides a feasible path to guide DRL early-stage exploration, a DRL module for channel guidance encourages a robot to move along a channel and guides the robot to bypass an obstacle, the path length of a robot operation space is optimized, and the path length of the robot operation space is optimized; an inverse solution module fusing a greedy algorithm and a neural network calculates loss according to the target pose and the current pose predicted by the neural network, the input of the neural network is adjusted in the gradient descent direction through a greedy strategy, and when the neural network tries to adjust the near-end joint of the robot to be free of fruit, the neural network is started; and then trying to adjust the near-end joint and the adjacent joints of the robot until the path of the robot operation space is converted into the path of the joint space, so that the robot path planning task is completed. The RRT-star module overcomes the problem that pure reinforcement learning is low in exploration efficiency in a complex environment, the channel-guided DRL module reduces collision between the robot and an obstacle, the inversion solution module fusing the greedy algorithm and the neural network expands and adjusts adjacent joints when the near-end joints are preferentially adjusted and necessary, and calculation redundancy is reduced.

The present invention discloses a transmitting beamforming method for a phased array system based on the Riemannian Newton method. This method transforms the problem of minimizing the beam pattern matching error under constant modulus constraints into an unconstrained optimization problem in the complex circular manifold, and uses the second-order Riemannian Newton method to solve it, so as to ensure that the direction updated in each iteration is always the descending direction of the objective function and has a superlinear convergence rate. The present invention has better performance advantages in terms of main lobe fitting accuracy and side lobe level reduction, and has a faster convergence rate.

This invention belongs to the field of image generation technology, specifically relating to a semantic alignment method and system for image generation based on a diffusion model with dual-loss optimization. The method includes: extracting cross-attention maps corresponding to each topic word in the text prompt during denoising; calculating aggregation loss and isolation loss based on the cross-attention maps, whereby the aggregation loss aggregates attention activations corresponding to the same topic word, and the isolation loss separates attention activations corresponding to different topic words; constructing a joint loss function based on the aggregation loss and isolation loss, and updating the latent representation along the gradient descent direction; inputting the updated latent representation into the denoising process at the next time step, until a preset number of denoising steps is reached, generating the output image. This invention optimizes the aggregation loss and isolation loss, simultaneously solving the problems of internal scattering and overlap in the attention maps, and can be applied to downstream tasks such as controllable layout generation and personalized generation.

The invention discloses a cross-domain cluster unmanned collaborative lotus root picking and collecting method and system. According to the method, a map is sequentially executed in blocks, that is, global path planning is converted into local path planning, and a main task is decomposed into a plurality of sub-tasks; in each sub-task, firstly, position information of lotus roots and obstacles is determined through an existing detection and recognition method, then, based on an artificial potential field method, a gravitational field is set for the positions of the lotus roots, a repulsive force field is set for the positions of the obstacles, and then the lotus roots are gradually moved to the positions of the lotus roots by bypassing the obstacles in the total potential field gradient descending direction to be dug; and the purpose of detecting and identifying at the same time is achieved. The sub-tasks are executed in sequence, and the lotus root digging task is completed. According to the method, the whole lotus root field can be traversed comprehensively, the path can be adjusted adaptively according to real-time sensor data and environment changes, so that collision with obstacles is avoided, smooth completion of tasks is ensured, the coverage is large, and the robustness is good.

This application provides a point cloud processing system and method. The system includes: a PL (Plug-in) terminal and a PS (Power Supply) terminal. The PS terminal sends point cloud data calculation commands to the PL terminal. A hash filtering module generates point cloud data read commands based on the point cloud data calculation commands, sends the point cloud data read commands to a data transfer module, and processes the initial point cloud data in real time using a pipeline to obtain processed point cloud data. A gradient calculation module performs gradient calculations on the processed point cloud data in parallel using a data stream through multiple calculation sub-modules connected by FIFO or ping-pong buffers to obtain multiple gradient vectors. An application processor on the PS terminal calculates the descent direction of the multiple gradient vectors and determines whether the descent direction converges. If the descent direction converges, the point cloud computing is terminated. Thus, the provided point cloud processing scheme can improve the throughput of point cloud data processing and reduce latency.

The invention discloses a load flow calculation model and method based on track joint optimization and physical information, and belongs to the technical field of electric power. The model comprises a KAN-based physical information model and a trajectory joint optimization algorithm; comprising the following steps: applying an interpretable KAN to load flow calculation; designing a physical constraint loss function to realize embedding of a physical mechanism; adopting a trajectory joint optimization algorithm to convert the training of the KAN into a dynamic system evolution problem; the physical information model based on the KAN is of a double-layer architecture, the first layer calculates a node voltage amplitude, the second layer outputs a voltage phase angle, the physical constraint network layer calculates node power through a power flow network, and an output result is constrained to meet a power balance equation through residual calculation; according to the trajectory joint optimization algorithm, parameters are guided to evolve and converge to an optimal solution along an energy descending direction by constructing a nonlinear dynamic system. According to the invention, the limitation of a black box network can be broken through, the problem of large-scale KAN network training challenge is solved, and the precision is improved.

A path planning method based on analytical gradients of a neural network includes: acquiring sample datasets of a rocket under multiple typical flight conditions; training the neural network with flight control variables as input and aerodynamic coefficients as output until convergence is met, resulting in a surrogate model for fitting the rocket's aerodynamic characteristics; collecting the rocket's current state vector and control vector, inputting the control vector into the surrogate model to obtain the current aerodynamic coefficients, and constructing the rocket's flight state equation; based on the flight state equation, using an automatic differentiation method, analytically calculating the gradient with respect to the control vector to form a gradient sensitivity matrix; using the gradient sensitivity matrix to determine the descent direction; and iteratively updating the control vector using a numerical optimization algorithm until the termination condition is met, outputting the optimal flight path. The optimal flight path obtained by this method more closely matches the rocket's actual aerodynamic environment, exhibiting higher terminal accuracy.

This invention relates to the field of surveying and mapping technology, and proposes a contour gradient checking method, system, and device, comprising: acquiring all contour line entities and their corresponding contour line information; determining whether any contour line entity and its information have been stored in a cache unit; if not, searching for adjacent contour lines along the gradient descent direction of the contour line entity, and storing the matched adjacent contour lines and their information in the cache unit; searching for the upper and lower contour lines of the contour line entity according to the cache unit storing the contour line entity and its information, or the corresponding cache unit of the contour line entity; if upper and lower contour lines exist, comparing the contour interval of the adjacent upper and lower contour lines with a set value, and storing the comparison result; if no upper and lower contour lines exist, checking the next contour line entity; repeating the above steps until all contour line entities have been checked; and displaying the comparison result of the upper and lower contour lines of the target contour line entity.

The present invention discloses a method for matching the transmitting direction pattern of a MIMO array based on a Riemann complex circular manifold, belonging to the technical field of radar. In the present invention, the waveform is designed to match the ideal transmitting direction pattern. Assuming that there already exists an ideal transmitting direction pattern and the detected airspace range, the problem is modeled as minimizing the squared difference between the ideal transmitting direction pattern and the designed transmitting direction pattern, satisfying the constant modulus constraint of the waveform, and the Riemann complex circular manifold method is adopted to effectively solve the constant modulus constraint problem in the design of the MIMO array transmitting waveform: First, the Euler gradient is calculated, and the Riemannian gradient is calculated based on the Euler gradient, and then the descent direction is updated. Finally, the feasible solution is updated on the Riemann complex circular manifold to obtain a transmitting waveform that matches the ideal direction pattern. The present invention can obtain better matching performance and sidelobe values in the transmitting direction pattern, which is more conducive to improving the detection performance of the MIMO array for targets.

This disclosure provides an integrated clock gate placement method, apparatus, device, and storage medium based on gradient descent. The method includes: determining the anchor point of the integrated clock gate based on the register position of the register controlled by the integrated clock gate; determining a corresponding line length loss function based on a first distance between the integrated clock gate and the anchor point, and a second distance between the integrated clock gate and the register; a first weight of the first distance being greater than a second weight of the second distance; determining a total loss function based on the line length loss function and a placement loss function; the placement loss function including at least one of a density loss function, a congestion loss function, or a timing loss function; and updating the position of the integrated clock gate based on the gradient descent direction of the total loss function to obtain a placement position that satisfies the convergence condition of the total loss function. This method achieves multi-dimensional improvements in placement quality while ensuring clock tree performance.

The invention discloses a large language model zero-order fine tuning method and system based on low-rank projection matrix learning, and belongs to the technical field of model fine tuning, and the method comprises the steps: during low-rank projection matrix learning, freezing large language model parameters to be used for text processing, utilizing small-batch training sample data, updating a low-rank projection matrix through zero-order optimization, and carrying out zero-order fine tuning on the low-rank projection matrix; the low-rank projection matrix can be multiplied with the low-dimensional random matrix when the model parameters are updated, disturbance with the same dimension as the model parameters is constructed, and the disturbance direction in a low-rank disturbance space formed by the low-rank projection matrix can be close to the descending direction of the loss function by adjusting the low-rank disturbance space; and constructing a low-rank disturbance matrix by utilizing the low-rank projection matrix obtained by learning when model parameters are updated, and performing zero-order gradient estimation and updating on the model parameters of the large language model by combining all training sample data. Therefore, the text processing capability of the large language model is improved while the optimization efficiency in the high-dimensional parameter space is improved and the gradient variance is reduced.