12 results about "Minimum entropy" patented technology

The present application provides a method for bearing fault diagnosis across working conditions by using a minimum entropy optimized prototype contrast network, and relates to the technical field of intelligent fault diagnosis. In the pre-training stage, an auxiliary domain discriminator is constructed to assist DA with the discriminant information of the classifier, the classification difficulty is evaluated by sample entropy, and the performance degradation in the DA process is inhibited. In the training stage, the learning vector quantization method is adopted to find the prototype. Through intra-domain prototype contrast learning, the sample features are closely gathered around the same prototype in the feature space, while being separated from the different prototypes. Then, the intra-class consistency of the features is enhanced, and the inter-class distinguishability is improved, so as to realize the precise alignment in the feature space. In addition, the cross-domain instance-prototype learning aligns the semantic structure in the shared embedding space, alleviates the negative transfer problem through the fine-grained alignment strategy, and improves the generalization ability of the model. Through the pseudo-label generation and the weighted loss function, the generalization performance of the model in the cross-working-condition small sample scene is improved.

Technological advancements are disclosed that utilize inertial sensor data for multiple classes to select a combination of filters to extract information though features to train a machine learning core decision tree. A determination is made whether the data for a class includes a frequency peak or dominating frequency that contains significant information about the class. In response to the data for the class including a frequency peak, a peak-based frequency range is determined. An entropy value is calculated for multiple frequency ranges in the data for the class. An entropy-based frequency range is selected from the multiple frequency ranges having a minimum entropy value. A frequency of interest is selected from the peak-based frequency range and the entropy-based frequency range for the class. A combination of filters is selected for each frequency of interest for each class and a decision tree is trained based on selected filter combination.

The present application belongs to the technical field of information security and random number generator, and aims to solve the problem that the traditional minimum entropy estimation method does not consider the influence of environmental factors and other multi-source data on the quality of physical entropy source random number. A multi-source information fusion physical entropy source security evaluation method is provided, comprising the following steps: obtaining physical entropy source data and processing to obtain a multi-source fusion data set; performing feature extraction on the multi-source fusion data set through a variational autoencoder network, and outputting a latent representation as the data after feature extraction; inputting the latent representation divided into a test set into a verified hybrid deep learning neural network model to obtain a prediction result of the physical entropy source; calculating a global prediction probability and a local prediction probability based on the prediction result; obtaining a minimum entropy evaluation result based on the global prediction probability and the local prediction probability; and judging the security of the physical entropy source based on the minimum entropy evaluation result. The present application can improve the low accuracy of existing physical entropy source evaluation.

The invention relates to a self-adaptive quantum random number generation method based on real-time entropy estimation, and belongs to the technical field of quantum random number generation. According to an existing integrated quantum random number generator, due to the non-stationary characteristic of a physical entropy source, the entropy rate dynamically drifts, and after-processing parameters are fixed, randomness safety and generation efficiency are difficult to guarantee at the same time. According to the method, the state transition probability of the original random bit stream is counted in real time, the confidence interval correction is introduced to calculate the minimum entropy, the input bit width is dynamically adjusted accordingly, the Toeplitz matrix is reconstructed for random extraction, and self-adaptive adjustment of the compression ratio is achieved. According to the method, the compression ratio is increased to ensure the security of the output random number when the entropy source quality is reduced, and the compression ratio is reduced to improve the generation efficiency when the entropy source quality is improved, so that the dynamic balance between the randomness security and the generation efficiency is realized on the premise of keeping integration and low cost.

The invention belongs to the technical field of information security, and particularly relates to a random number security rapid evaluation method and device based on a neural network. According to the method, a random bit stream is mapped into a high-dimensional time state sequence by utilizing the high-speed nonlinear mapping capability of a photon reserve pool, and key time structure characteristics are highlighted by virtue of a time sequence mode attention mechanism, so that sensitive detection on random changes of a physical entropy source is realized. Compared with an existing method depending on a statistical test or a large-scale deep learning model, the method fully utilizes the high-speed calculation capability of the photon reserve pool neural network and the extraction capability of a time sequence mode attention mechanism on a key time structure, does not need complex model training, is light and efficient in calculation process, and is easy to implement. The method can be easily integrated with optical random number generator hardware, and is suitable for real-time minimum entropy evaluation of a high-speed physical entropy source.

This invention provides an RBAC+ABAC-supported access control method for AI knowledge bases, belonging to the field of AI knowledge base technology. This invention employs a gradient boosting decision tree and a Bloom filter to construct a simplified decision model. An optimization algorithm based on thermodynamic entropy finds the policy configuration with the minimum entropy value through simulated annealing, eliminating rule conflicts and simplifying the logical structure. During vector retrieval, the permission encoding vector and semantic vector are concatenated to form a joint vector to construct a permission-aware index. During answer generation, a permission context fingerprint is attached, and secondary propagation of permissions is verified through a permission re-verification gateway. Operation logs are recorded using a hash chain structure to ensure the immutability of audit traceability. This invention solves the technical problem that AI knowledge bases cannot simultaneously guarantee the efficiency of permission decisions and the maintainability of policy rules when performing fine-grained access control.

The invention discloses a diversity preserving supervision fine tuning method for reinforcement learning alignment, and particularly relates to the field of large language model supervision, comprising the following steps: step 1, constructing a supervision fine tuning model; 2, constructing a reward model; and step 3, optimization is carried out in combination with reinforcement learning. According to the method, the CE targets are redesigned or expanded, so that the CE targets can be kept aligned, and various comprehensive understanding can be explicitly encouraged to be reserved. Entropy regularization can prevent the model from collapsing into deterministic output, and a wider solution space is provided for subsequent RL. Variants of the method comprise adaptive entropy weight or a constraint form of forcing the minimum entropy level in training, and smooth cross entropy can prevent the model from completely collapsing into distribution, so that secondary hypothesis, comparison or boundary regularization is reserved, reference output advantages are guaranteed to be kept, meanwhile, a competition mode is not excessively suppressed, and the performance of the model is improved. And thus, diversity is kept on a local decision boundary.

The invention discloses a method of a semi-device independent quantum random number generator for resisting general attacks, and belongs to the technical field of continuous variable quantum random number generators. In order to solve the problem of strict requirements on a source end during resistance to general attacks in the prior art, the invention provides (1) a minimum entropy evaluation method which simultaneously considers general attacks of an inaccurately represented source and a measuring end, (2) three-input and multi-output configuration of a generator and (3) a heterodyne detection-based phase compensation method. According to the method, an energy boundary hypothesis and an Azuma inequality are combined, an authenticated true random number can still be generated when a source and measurement equipment are not accurately represented at the same time, and a limited code length effect is also considered. In the continuous variable generator, three-quantum-state input and multi-discrete-interval output are adopted, and the entropy generated in each round can be improved. Heterodyne detection allows phase compensation on post-processing of samples, avoids application of an active phase stabilization system, and reduces complexity of a generator system.

The invention relates to a recommendation model training method for applying minimum entropy bucketing to cross statistical features. The recommendation model training method comprises the steps of obtaining a feature information table of a user and an interaction behavior log of the user and a content set; generating a cross statistical characteristic value based on the interaction frequency of the two; constructing a training sample set based on the cross statistical feature values; performing recursive segmentation operation on the cross statistical feature value sequence in the training sample, and iteratively searching an optimal segmentation point to construct a bucket boundary by taking the minimum conditional entropy as a criterion in the recursive process; and mapping the cross statistical characteristic value in each training sample into a bucket number, inputting the bucket number and the corresponding label value into the recommendation model to be trained, and optimizing the parameters of the recommendation model to be trained by adopting a cross entropy loss function until the loss function value converges and outputting the trained recommendation model. According to the method, the loss function value of the model can be remarkably reduced, and key evaluation indexes such as click rate prediction accuracy are remarkably improved.

The invention provides a cross-working-condition bearing fault diagnosis method of a prototype comparison network with minimum entropy optimization, and relates to the technical field of intelligent fault diagnosis. In a pre-training stage, an auxiliary domain discriminator is constructed, DA is assisted by discrimination information of a classifier, classification difficulty is evaluated through sample entropy, and performance degradation in a DA process is inhibited; in the training stage, a prototype is found in a learning vector quantization mode; through intra-domain prototype comparative learning, sample features are prompted to closely gather around similar prototypes in a feature space, and meanwhile, the sample features are kept separated from heterogeneous prototypes; further, the intra-class consistency of the features is enhanced, the inter-class distinction degree is improved, and therefore intra-domain accurate alignment is achieved on the feature level; besides, the cross-domain instance-prototype learning aligns the semantic structure in the shared embedding space, and through a fine-grained alignment strategy, the negative migration problem is relieved, and the generalization ability of the model is improved; and the generalization performance of the model in a cross-working-condition small sample scene is improved through pseudo label generation and a weighted loss function.

The present application relates to a kind of mechanical fault diagnosis systems based on minimum entropy deconvolution and stochastic resonance, comprising: the vibration data of mechanical rotating component is collected as original signal;After frequency scaling, the original signal is input into bistable stochastic resonance system, and correlation kurtosis is used as index function, system parameters are adaptively adjusted, so that the desired signal is output, the numerical solution of nonlinear system output is obtained using 4-order Runge-Kutta algorithm, the impact component in original signal is amplified using stochastic resonance phenomenon, and the signal after noise reduction is obtained;The signal after noise reduction is filtered again using minimum entropy deconvolution, and minimum entropy deconvolution filter signal is obtained;The signal after minimum entropy deconvolution filtering is subjected to hilbert envelope spectrum analysis, and the fault of mechanical rotating component is diagnosed.The present application can autonomously extract the frequency of fault signal by designing corresponding index under the condition that the fault frequency of unknown signal is unknown, and accurate fault diagnosis and positioning are realized.

The invention belongs to the technical field of network security, and particularly relates to a security gateway access strategy optimization method and system based on a zero-trust architecture, and the method comprises the steps: collecting behavior information of an access subject, and constructing a space-time association diagram based on access behaviors; the quantitative evaluation of the resource sensitivity is realized by calculating the information entropy and the conditional entropy of resource access; a Brownian motion model is adopted to analyze drift characteristics of access behaviors, and an abnormal behavior identification baseline is established; modeling an access control process as an attack and defense game, and balancing security, overhead and complexity through a game revenue function; selecting an optimal strategy combination based on a minimum entropy increase principle; according to the method, the access strategy can be dynamically optimized according to the spatial-temporal characteristics of the access behavior, the complexity is effectively controlled while the system security is ensured, and the method has important theoretical value and practical value.