19 results about "Mean square" patented technology

The invention discloses a pilot optimization method for large-scale MIMO channel estimation based on structural compressed sensing. The method comprises the steps of establishing a channel estimation model for a large-scale MIMO-OFDM (Multiple-Input-Multiple-Output-Orthogonal Frequency Division Multiplexing) system when pilots are placed in an overlapping mode; simplifying the channel estimation model for the large-scale MIMO-OFDM system, thereby enabling the channel estimation model to correspond to a structural compressed sensing model; and obtaining an optimum pilot matrix through utilization of a pilot optimization algorithm. Through adoption of the optimum pilot matrix, according to the channel estimation of the large-scale MIMO system based on structural compressed sensing, the mean square errors MSEs of the channel estimation are clearly reduced, and the channel estimation performance is improved.

The application relates to the technical field of high-dimensional data, in particular to a high-dimensional data fitting algorithm and system based on an improved gradient descent, the algorithm is deeply coupled with a long short-term memory (LSTM) network, a gated differential adaptive momentum variable gradient descent (AM-VGD) is designed for high-dimensional time series data, high-precision fitting of high-dimensional data is realized, the gated differential adaptive momentum variable gradient descent (AM-VGD) is proposed, different momentum factors and gradient decoupling items are designed according to the gradient characteristics of the LSTM forget gate, input gate and output gate, the gradient propagation efficiency is improved by 40% under high dimension, the model convergence iteration number is reduced to <=500 times, the convergence speed is improved by more than 2 times, a stacked auto-encoder (SAE) is fused to decouple and extract high-dimensional features, a high-dimensional decoupling regular term of the gradient descent is combined, the collinearity interference between features is eliminated, the fitting determination coefficient R2 of high-dimensional time series data is greater than or equal to 0.98, the mean square error (MSE) is reduced by more than 60%, and there is no overfitting / underfitting phenomenon.

The application discloses a two-dimensional semiconductor device quantum transport simulation method based on machine learning optimization of kp parameters. The method comprises the following steps: obtaining crystal structure information of a target two-dimensional material, obtaining band structure data, and performing standardization processing; obtaining crystal structure symmetry of the target two-dimensional material, and obtaining energy eigenvalues; constructing a loss function based on mean square error, and obtaining a parameter set that minimizes the loss function; based on a new two-dimensional material, obtaining a band structure data set, inputting the data set into a kp model Hamiltonian, and optimizing the parameter set, so as to obtain an optimal parameter set; applying the optimal parameter set to quantum transport simulation, calculating current-voltage characteristics of the target two-dimensional semiconductor device, and completing simulation of quantum transport of the two-dimensional semiconductor device. Through the application, a device simulation technology that is specially used for the target two-dimensional material and is efficient in calculation like traditional TCAD software can be realized while the calculation accuracy close to the first principle calculation is maintained.

The present disclosure relates to a model training method and device and a readable storage medium, wherein the method performs knowledge distillation on each first intermediate subnetwork of a teacher model, thereby avoiding the selection and setting problems of the distillation position and the number of distillation layers, and achieving better distillation effect in the intermediate subnetwork distillation task. In addition, the knowledge distillation is performed by using the boundary mean square error loss, which not only enables the student model to learn the output of the teacher model, but also enables the student model to learn whether the teacher model activates the output, so that the student model learns more knowledge and is beneficial to improve the performance of the target student model.

The application discloses a movable array element anti-interference position beam joint method facing unknown direction, introduces movable array elements, and makes array geometry structure be treated as adjustable optimization variable in deployment stage. By changing equivalent array element spacing and pattern structure of the array, the separable degree of expected signal and interference signal incident direction on the spatial response is improved, and the destructive influence of strong interference on the array output when falling in the high side lobe / grating lobe direction of the array is reduced. In each candidate position, an adaptive beam forming process with a pilot sequence as a reference is adopted to directly generate a beam forming weight vector corresponding to the position, so that the mean square error of the receiving output in the pilot interval is minimized. The "movable array element position-performance index" is modeled as a target function to be optimized, and a global adaptive search method based on a probability model is adopted to select the next evaluation position in an iterative manner, instead of exhausting evaluation on all candidate positions.

This disclosure relates to the field of bridge engineering, specifically to a design method, system, and medium for HPC-RC combined eccentrically compressed columns. The method includes: defining a design space; selecting optimal design data combinations based on engineering specification constraints, bearing capacity constraints, and cost-effectiveness values ​​to construct a design data sample set; constructing a two-branch heterogeneous neural network model, the model including an input layer, a shared feature extraction layer, a diameter prediction classification branch, and a reinforcement area prediction regression branch; constructing a loss function composed of focal loss and mean square error loss; training the two-branch heterogeneous neural network model using the design data sample set; and using the trained two-branch heterogeneous neural network model to predict the diameter and total reinforcement area of ​​the eccentrically compressed column. This disclosure improves the stability and accuracy of the design, reduces the computational burden, increases computational speed, and reduces computational complexity.

This invention provides a reference-free background calibration method for ultra-high-speed TIADCs based on minimum mean square (RMS) calibration. It employs a decimation-calibration-interpolation calibration mode. The original quantized data is decimated using a specific decimation rate to obtain multiple quantization code sets. A step-by-step grouping calibration algorithm is used based on the number of interleaved channels, selecting the first quantization code set for step-by-step calibration. Each step involves grouping the data and using a dual-channel synchronous detection method based on a binary search to search for the clock jitter value of each channel in the grouped quantized data. Other quantization code sets are then compensated using a first-order Taylor approximation in the digital domain based on the clock jitter value of the first quantization code set. Finally, the compensated multiple quantization code sets are re-interpolated to obtain the calibrated result of the original data. This invention eliminates the need for a reference channel, minimizing the RMS of the detected clock jitter value and thus improving compensation accuracy.

The application discloses a convex or approximately convex system timing modeling method based on an input convex Transformer architecture, mainly relates to the technical field of Transformer architecture, and aims to solve the problems that a high-precision timing model is difficult to be used for convex optimization control and a convex optimization friendly model is difficult to achieve high prediction accuracy. The method comprises the following steps: adding cross attention output and decoding hidden state to obtain an intermediate hidden state; inputting the intermediate hidden state into a convex neural network, performing preset layer recursion, and outputting a correction term; adding the intermediate hidden state and the correction term to obtain decoder output; using the decoder output, adopting non-negative output projection to obtain future state prediction of a next moment, and then obtaining a future state prediction sequence; based on the corresponding mean square error loss of the future state prediction sequence and a target state sequence, training and updating parameters of a historical encoder and a convex decoder, and then obtaining a trained historical encoder and a convex decoder.

This invention belongs to the field of artificial intelligence and model compression technology, and discloses an additive quantization method for large language models. Without changing the additive quantization encoding format and codebook structure, a sampling strategy that jointly considers sub-vector energy density and spatial coverage is introduced in the residual K-means initialization stage. This allows high-energy-density regions to obtain more reasonable codeword coverage, significantly improving the codebook initialization quality. In the block-by-block end-to-end fine-tuning stage, a direction alignment loss based on negative log-cosine similarity is introduced in addition to the mean square error. This explicitly constrains the directional consistency between the quantized block output and the full-precision block output, making the direction constraint and amplitude constraint complementary. This method requires no quantization-aware training or large-scale fine-tuning, does not introduce additional trainable parameters, has low computational and storage overhead, is simple to implement in engineering, maintains near-full-precision inference performance under extremely low 2-bit quantization conditions, and significantly reduces memory usage and energy consumption during model deployment.

Embodiments of the present application provide a method and device for constitutive model construction and parameter optimization of honeycomb materials, computer equipment and medium, relating to the technical field of electronic digital data processing, wherein the method comprises the following steps: obtaining original experimental data sets generated by mechanical experiments of the honeycomb material under multiple different working conditions; fitting to obtain a shape function representing the stress-strain relationship under a reference working condition, constructing a density function representing the influence of density on mechanical properties, using an exponential nonlinear function, constructing a strain rate function representing the influence of strain rate on mechanical properties, and constructing a temperature function representing the influence of temperature on mechanical properties; constructing a constitutive model of the honeycomb material based on multi-factor correction factors; taking the minimum mean square error between the stress-strain curve predicted by the constitutive model and the target curve as the optimization target, performing global iterative optimization and outputting. Through the constructed constitutive model, the problem of strong dependence of traditional model parameter fitting and poor universality is solved.

The present disclosure relates to the field of bridge engineering, and particularly relates to a design method, system and medium for HPC-RC combined eccentric compression column. The method comprises the following steps: defining a design space; screening out an optimal design data combination according to engineering specification constraints, bearing capacity constraints and performance-price ratio value, and constructing a design data sample set; constructing a double-branch heterogeneous neural network model, wherein the model comprises an input layer, a shared feature extraction layer, a diameter prediction classification branch and a reinforcement area prediction regression branch; constructing a loss function composed of focal loss and mean square error loss, and training the double-branch heterogeneous neural network model by using the design data sample set; and predicting the diameter of the eccentric compression column and the total reinforcement area of the eccentric compression column by using the trained double-branch heterogeneous neural network model. The present disclosure improves the stability and accuracy of the design, has a small calculation burden, improves the calculation speed and reduces the calculation amount.

This invention discloses a synchronization method for a semi-Markov jumping neural network with measurement loss, belonging to the field of synchronization control technology for semi-Markov jumping neural networks. This invention can guarantee the mean-square global asymptotic stability of the error system when external disturbances are zero, and meet a preset disturbance attenuation level when external disturbances exist. The controller in this invention explicitly includes a non-fragile term structure, covering the control gain to realize error / uncertainty, avoiding the risk of "nominal design feasibility but implementation failure." This invention employs a dynamic event triggering mechanism to avoid Zeno risk at the mechanism level through "minimum trigger interval," and avoids performance degradation caused by long-term lack of updates through "maximum interval forced update." Dynamic variables achieve adaptive trigger thresholds, thereby reducing unnecessary transmissions while meeting performance requirements.

This invention discloses a method and apparatus for table anomaly detection to improve the security of network platforms. The method includes: feeding the original feature matrix into a Transformer branch, capturing global cross-column dependencies using multi-head self-attention, and injecting relative position information using rotational position encoding to output global features; using the global features output from the Transformer branch as input to a depthwise separable convolutional encoder, extracting local neighborhood features through channel-wise and point-wise convolution to obtain global-local fused features, which are then used as input to initiate parallel dual-branch operations; using the anomaly score obtained from the reconstruction error of the depthwise separable convolutional decoder through a learnable scaling layer as a soft label, calculating the mean square error with the anomaly probability obtained from the classification branch through softmax, forming a consistency regularization term; weighting and summing the reconstruction loss, classification loss, and consistency loss according to learnable weights to obtain the overall loss, and training the entire network end-to-end through gradient backpropagation; and automatically optimizing the network structure hyperparameters using Bayesian optimization to finally obtain the table data anomaly detection results.

The application discloses a geodesic line learning method based on Riemann geometry prior, and belongs to the technical field of machine learning. P，gp (s) the decoder of the trained autoencoder, calculate the Riemann metric g and Riemann connection e of all sampling points, bring all geodesic equation values into the mean square error loss function, and minimize the difference between the geodesic equation values and 0. The application takes the geodesic equation as a geometric prior, adjusts the Bezier curve control point set for generating the geodesic path, has better robustness when facing noise or unevenly distributed high-dimensional data, and can more accurately calculate the geodesic path.

The present application belongs to the technical field of power quality measurement, and provides a harmonic distortion type identification method and system based on symmetry characteristics, which obtains power signal data according to a preset sampling window sliding; processes the obtained data, extracts harmonic components, calculates single-window harmonic energy variance indexes, performs grouping processing, compares with a preset center-symmetry normalized mean square error threshold, judges whether each group of data conforms to center-symmetry distribution, and then compares with an error threshold of inter-harmonic characteristics to obtain an inter-harmonic criterion variable; processes, calculates and analyzes time domain data of any complete single period, compares the consistency degree of the translated first half cycle data and the inverted second half cycle data to obtain an even harmonic criterion variable; and comprehensively identifies the signal harmonic distortion type according to the inter-harmonic criterion variable and the even harmonic criterion variable. The present application realizes accurate identification of the power signal harmonic distortion type.

This invention discloses a rapid geometric modeling method and apparatus. The method includes extracting the three-dimensional symbolic distance field information of all frames of a human body template action sequence and processing it to obtain a data matrix; extracting the human body template torso skeleton of the first frame of human action; initializing the data matrix with parameters based on human body parts according to the human body template torso skeleton; inputting the parameters into a reconstruction model based on a quadratic kernel gate function to obtain the reconstructed symbolic distance fields of all frames; optimizing and updating the model parameters using the expectation-maximization algorithm; calculating the mean square error of the reconstructed symbolic distance fields of all frames and the original symbolic distance fields; reconstructing the symbolic distance field of each frame using the parameters corresponding to the minimum mean square error value; and reconstructing the three-dimensional mesh of the human body template from the reconstructed symbolic distance fields of all frames. This invention can effectively model the dynamic geometry of the three-dimensional human body, enabling the understanding of human actions to have spatiotemporal inductive generalization.

This invention proposes a reconfigurable metasurface beamforming design method based on CGAN and Gumbel-Sinkhorn. This method constructs a conditional generative adversarial network (GAN) as the inverse design network. The generator takes the target beam pattern as a conditional input to generate the corresponding encoding matrix. The Gumbel-Sinkhorn operator is introduced to discretize this continuous matrix, transforming it into an approximately doubly random permutation matrix. The encoding is sorted and selected by multiplying this permutation matrix with a predefined structure vector (SV), outputting a discrete 0 / 1 metasurface encoding array. A noisy forward prediction network is trained to predict the beam pattern corresponding to a given encoding array. In the training of the inverse design network, the discretized encoding matrix output by the generator is input into the noisy-trained forward prediction network. The mean square error between the predicted beam pattern and the target beam pattern is calculated as the target loss term to optimize the network parameters and accurately learn the complex mapping relationship from the beam pattern to the encoding array.

A first calibrator detects mismatches in the capacitor array and updates a look-up table (LUT) with calibrated weights that are copied to a positive LUT and a negative LUT, which are then adjusted for non-linear errors by a second calibrator using a least mean square (LMS) algorithm. The binary code in a successive approximation register (SAR) is complemented to generate a complement with a sign bit. When the sign bit is positive, entries from the positive LUT with complemented data bits = 1 are read and summed, a first offset is added, and the sum is normalized to obtain a corrected code. When the sign bit is negative, entries from the negative LUT with complemented data bits = 0 are read and summed, a second offset is added, and the sum is normalized to obtain a corrected code. A multivariate stochastic gradient descent generates polynomial coefficients that further correct the corrected code.

The application provides a kind of tunneling parameter multi-step prediction method for open type tunnel boring machine.The method is applied to TBM tunneling technical field, and includes collecting tunneling parameters such as boring machine advancing speed, cutterhead rotating speed, cutterhead torque, total advancing force, main machine belt machine speed, etc.; based on cutterhead torque and advancing speed, effective tunneling state is determined by time sequence logic, effective tunneling section data is extracted and numerical validity is cleaned, and a multi-step prediction time sequence training set is constructed according to preset input step span N and output step span M; a deep prediction model is constructed by fusing LSTM and large model architecture, trained based on the training set, and the optimal hyperparameters are determined by mean square error loss function and grid search to obtain the final prediction model; input the tunneling parameters to be predicted, and output the prediction results of the next M time steps. Thus, the accuracy of the multi-step prediction of the tunneling parameters of the open type tunnel boring machine is improved.