80 results about "Dichotomic search" patented technology

The invention relates to the technical field of graphics card power consumption adjustment, and discloses a method, a device and equipment for dynamically adjusting the power consumption of a graphics card component. The method comprises the following steps: carrying out multi-channel data acquisition and working load characteristic analysis on operation parameters of a graphics card GPU core, a video memory and a power management unit to obtain a component temperature state evaluation matrix; executing component-level power consumption distribution based on multi-objective optimization to obtain a power consumption distribution scheme of each component; inputting the power consumption distribution scheme of each component into a power consumption-frequency-voltage mapping model, and calculating an optimal frequency voltage configuration parameter through a binary search algorithm to obtain a frequency voltage control instruction; performing multi-step prediction to obtain a module power consumption demand prediction value; according to the method, the frequency voltage combination with the highest performance efficiency can be found under the condition that the power consumption constraint is met, and accurate power consumption control is achieved.

The invention discloses a recalculation perception model splitting method and system oriented to assembly line parallel training, and belongs to the field of artificial intelligence of computer science. The method specifically comprises the steps that a dijkstra algorithm and two-layer binary search are used, and the minimum value of the sum of forward propagation time and reverse propagation time of assembly line microbatch under re-calculation perception model splitting is calculated; and an intermediate result obtained by a dijkstra algorithm in a search process is used to calculate and output an approximately optimal model splitting strategy for re-calculation perception. According to the method, the model splitting method and system are designed by utilizing the model characteristics of a Tasformer large model and the characteristics of an assembly line parallel mode, the model is non-uniformly split to each stage of the assembly line, so that the calculation load of each stage of the assembly line is balanced when re-calculation is started, and the calculation efficiency is improved. Therefore, the overall performance of large model pipeline parallel training and the utilization efficiency of cluster equipment are improved.

The invention discloses an intelligent literature screening method and system based on zero sample voters. The method comprises the following steps: expanding a literature set based on seed literatures, constructing a classification system comprising a plurality of zero sample voters, screening literatures related to a target topic through a majority voting mechanism, and iteratively expanding the literatures to obtain more literatures related to the target topic. The zero sample voting device integrates subject term matching, language model pre-training, multi-large language model voting and a semantic embedding vector technology, adaptively determines a threshold value in combination with an improved binary search algorithm, and realizes high-precision classification without training data. The system supports parallel processing and intelligent API management, and the processing efficiency of a large-scale literature set is remarkably improved. The method is suitable for the fields of academic research, literature retrieval and the like, can quickly construct a large-scale high-correlation literature information base, and particularly meets the retrieval requirements of emerging technologies or interdisciplinary themes.

An unmanned aerial vehicle assisted MEC network energy efficiency optimization method based on reinforcement learning belongs to the field of mobile communication, and comprises the following steps: firstly, carrying out system modeling and state information collection, inputting state information to a strategy model, generating flight and energy transmission actions, and carrying out unmanned aerial vehicle movement and service equipment selection; an energy transmission and task unloading model is defined, and a gradient-based equipment unloading time distribution algorithm is started; then initializing binary search of the equalization marginal utility gradient, executing single iterative search of the equalization marginal utility gradient, updating a search interval of the equalization marginal utility gradient, and completing search of the equalization marginal utility gradient; then determining a final time allocation scheme, executing energy transmission and task unloading, collecting performance data, calculating a reward function and storing experience; and finally, updating the Critic network, updating the Actor network and softly updating the target network, and obtaining an optimal strategy after iterative optimization. The overall energy efficiency and task processing throughput of the system are remarkably improved.

The invention provides a cellular-free mMIMO (multiple input multiple output) system computing power resource allocation and beam forming optimization method and a cellular-free mMIMO system computing power resource allocation and beam forming optimization system, and belongs to the technical field of wireless communication. The method specifically comprises the following steps of: respectively establishing a double-queue model based on a first-in first-out criterion at a central processing unit (CPU) and an access point (APs), taking a dequeue rate at the CPU as an enqueue rate at the AP, constructing a data volume relationship in a queue of a current frame and a previous frame, and analyzing system downlink end-to-end time delay based on an M / D / 1 queuing model. In a first layer queue, a binary search algorithm is provided to design computing power resource allocation at a CPU, and in a second layer queue, a path tracking algorithm is provided to optimize a beam forming vector at an AP. According to the method, the instantaneous channel state information is utilized, the worst user time delay is minimized under an alternating iteration algorithm framework, the calculation complexity is reduced, and meanwhile an effective solution is provided for ultra-low time delay communication.

The invention discloses a zero sample target detection algorithm based on binary search and CLIP, and belongs to the technical field of image target detection, and the algorithm specifically comprises the steps: constructing a data set training CLIP model, carrying out the preprocessing of a target image, setting an initial range of boundary search, inputting the preprocessed target image into the CLIP model, and carrying out the recognition of the target image. The method comprises the following steps: determining a primary bounding box of a target by using binary search and a pad method, calculating the similarity between a target image and a target text by using a CLIP model, generating text description information of the target, adjusting and optimizing the primary bounding box in combination with position information of the target in the image, and generating a final bounding box. The method has remarkable advantages in the aspects of processing new category target detection, improving the calculation efficiency and reducing the manual annotation cost.

The invention discloses a DCT (Discrete Cosine Transform) frequency domain watermark embedding optimization method and system based on a potential space, and the method comprises the steps: firstly coding an original image into the potential space, segmenting the original image into small blocks, embedding a single-bit watermark into the frequency domain of each small block through weighted voting, then carrying out the inverse transformation to a spatial domain, merging, and forming a potential representation with a watermark; decoding through a stable diffusion model to generate an image with a watermark; iteratively optimizing the potential representation through a multi-objective loss function until the loss is minimum; and finally, decoding the optimized representation again, linearly mixing with the original image, adjusting a mixing coefficient through binary search, and outputting a final image on the premise of meeting preset image quality. According to the method, a multi-dimensional robustness optimization strategy is combined, the stability and the non-erasable property of the watermark under the attack of the generative model are systematically improved, and the core requirements of AI copyright protection and content traceability are met.

This invention relates to the field of multi-processor multi-task joint scheduling technology, specifically disclosing a typed task joint scheduling system and method based on a heterogeneous multi-core architecture. It introduces a joint scheduling mechanism to divide and sort tasks according to task size and priority. Based on this, for the time-limited characteristics of real-time tasks, an inertial weight coefficient particle swarm optimization method is used to iteratively update the load balancing strategy, minimizing the maximum response time of real-time tasks while meeting their schedulability requirements. Simultaneously, for the low-priority characteristics of non-real-time tasks, the problem-solving is simplified using a Lagrange-based convex optimization approach, and an energy-constrained binary search algorithm is employed to effectively reduce the average response time of non-real-time tasks, achieving optimal load distribution for the system.

The invention discloses an attack resisting method and device for a prediction model, and belongs to the field of artificial intelligence security. And indirect targeted countermeasure attacks are realized aiming at the network range flight delay prediction model. According to the method, significance of non-target nodes is evaluated based on gradient, then a preliminary candidate range including a first damaged node subset is determined by using binary search external circulation and nesting internal circulation, and on this basis, gradual refining of the external circulation and nesting of the internal circulation are performed to approach a minimum damaged node subset. In the two circulation processes, the intersection of the first damaged subset and the second damaged subset of two adjacent external iterations is accumulated to obtain a second initial sample and a third initial sample, disturbance information is fully utilized, the antagonistic potential of damaged nodes is maximized, the number of the damaged nodes needed by attacks is reduced, and the attack efficiency is improved. The minimization capability of the victim node subset is effectively improved, and the attack budget is saved.

The invention provides an industrial edge network micro-service optimization method based on dynamic arrangement and migration expansion, which comprises the following steps: abstracting an industrial edge network into an intelligent industrial network model formed by heterogeneous edge nodes and links, modeling an industrial application request into a request chain formed by combining a plurality of micro-services with a data dependency relationship in sequence, calculating a micro-service instance deployment scheme and a request routing scheme; counting a time-varying request stream, analyzing the minimum core number meeting the time delay constraint in a statistical result by using a multivariate performance analysis model based on an m / m / c queue network and a binary search algorithm, and carrying out dynamic expansion and contraction adjustment on core resources of deployed micro-service instances; and solving a mixed integer nonlinear programming migration decision problem according to the initial dynamic scaling adjustment result and the migration cost so as to optimize the optimal collaborative deployment and request routing of the micro-service instance between the edge nodes. According to the invention, the service continuity and the overall energy efficiency of the industrial edge network system are improved.

The invention discloses a learning and optimization-based power grid interaction type residential building online cooperative operation method, which belongs to the technical field of power grid interaction type residential building cooperative operation and artificial intelligence crossing, and comprises the following steps of: establishing a problem of minimizing the operation cost of a single residential building during a non-power grid service period; modeling a single residential building operation cost minimization problem as a Markov decision process; training a residential building energy management agent by adopting a cross normalization algorithm to obtain an energy management strategy; total power limitation set by a power distribution system operator is considered, and a resident building group double-layer collaborative operation cost minimization problem is established; solving the collaborative operation cost minimization problem based on the energy management strategy obtained by each residential building and optimization auxiliary binary search, and executing a decision; the power distribution system operator calculates economic compensation and issues the economic compensation according to the energy management decision of each residential building before and after collaboration, the energy cost can be effectively saved, and the higher power grid service capability is achieved.

The invention discloses a method and a device for constructing an open-circuit voltage-state-of-charge model of a battery, and relates to the technical field of battery management systems. According to the method, when the OCV-SOC model of the battery is constructed based on the SVR model, the initial search space of the to-be-optimized parameters of each dimension of the SVR model is subjected to continuous binary segmentation and binary search, so that the optimal search subspace can be quickly positioned from the large-range initial search space of the to-be-optimized parameters, and the optimal search subspace can be quickly searched. According to the OCV-SOC model construction method, the parameter performance of each search subspace in the wide-range initial search space can be comprehensively evaluated through the binary segmentation and binary search of the initial search space, so that the OCV-SOC model construction method has the advantages that the parameter performance of each search subspace in the wide-range initial search space can be comprehensively evaluated, and the OCV-SOC model construction efficiency can be improved. According to the method, the target search space of the PSO algorithm can be accurately determined, the possibility of losing optimal parameters is avoided, and the accuracy of OCV-SOC model construction is improved.

A reinforcement learning-based energy efficiency optimization method for UAV-assisted MEC networks, belonging to the field of mobile communication, is proposed. First, the system is modeled and state information is collected. The state information is input into the policy model to generate flight and energy transfer actions, enabling UAV movement and service device selection. Next, an energy transfer and task offloading model is defined, and a gradient-based device offloading time allocation algorithm is initiated. Then, a binary search for the equilibrium marginal utility gradient is initialized, followed by a single iterative search, updating the search interval, and completing the equilibrium marginal utility gradient search. The final time allocation scheme is then determined, energy transfer and task offloading are executed, performance data is collected, the reward function is calculated, and experience is stored. Finally, the Critic network, Actor network, and target network are updated, and the optimal policy is obtained through iterative optimization. This invention significantly improves the overall energy efficiency and task processing throughput of the system.

This invention belongs to the field of radiation source individual identification technology, specifically relating to an open-set individual identification method based on closed-set model screening and binary search. This method leverages the intra-class clustering characteristics formed after training the closed-set model. First, it performs rapid screening of known classes on the test samples, reducing computational complexity while maintaining identification accuracy. Then, it adaptively determines the optimal discrimination threshold through a binary search strategy based on known class performance constraints, achieving the minimum missed detection probability while meeting engineering accuracy requirements. This significantly improves the system's open-set identification capability and real-time performance.

The application discloses a solar pumping system photovoltaic array capacity optimization method and device, relates to the field of new energy, and is used for minimizing photovoltaic array installed capacity design and energy storage capacity reduction. By acquiring basic parameters of a solar pumping system, the basic photovoltaic array installed capacity is obtained based on the peak water power of the system corrected based on the basic parameters. Under the condition of meeting the daily water supply constraint, the minimum feasible photovoltaic array installed capacity and the corresponding energy storage configuration are calculated based on the basic photovoltaic array installed capacity by using a dichotomy search method, with the minimum photovoltaic array installed capacity and energy storage capacity as the target. The application has the advantages of simple process, strong pertinence, robust optimization result, easy deployment, and strong expansibility.

The invention relates to a target allocation method based on binary search and a KM algorithm, and belongs to the field of resource allocation. According to the method, a target allocation problem is converted into a planning model, and an optimization function for minimizing a maximum value problem is determined; a bipartite graph is constructed, and in the bipartite graph, an optimization function is converted into a mode that in all distribution schemes, a maximum matching is solved, so that the maximum weight in the distribution scheme is the minimum, namely, a double-optimization problem is achieved; a solution space in a target allocation problem is determined, and then a target-resource matching scheme of maximum weight matching is obtained based on a target allocation method of binary search and KM algorithm fusion. The invention overcomes the limitation of the traditional method, and provides an efficient, accurate and flexible target allocation solution.

The invention discloses a method for detecting a visual area of a super-large scale intelligent super-surface (RIS) auxiliary communication system, the RIS is a uniform planar array, and the method comprises the following steps: sequentially detecting the first unit, the second unit and the fifth unit on an RIS array plane, and judging whether the first unit, the second unit and the fifth unit are shielded or not; delimiting an initial area according to the shielding conditions of the five units, and judging whether the initial area is completely shielded or not; if the initial area is completely shielded, a DFS algorithm is used for all units on the boundary of the initial area in sequence, all shielded units are detected, and therefore the final visible area is recovered; if the initial area is not completely shielded, a binary search strategy is adopted, a shielded unit is detected firstly, then the unit serves as a starting point, all shielded units are detected through a DFS algorithm, and then the final visible area is recovered. According to the invention, accurate visual area detection is realized with low complexity and measurement overhead.

The invention discloses a digital low dropout linear regulator based on improved successive approximation logic, which comprises a comparator, an MOS (Metal Oxide Semiconductor) transistor array, an improved successive approximation logic control module, a sampling-holding dynamic comparator, an analog-to-digital converter and a transient detector, a PMOS (P-channel Metal Oxide Semiconductor) transistor array is arranged in the analog-to-digital converter and has a comparison state and a PMOS transistor array state; the improved successive approximation logic control module is used for controlling the state switching of the analog-to-digital converter, whether the sampling-maintaining dynamic comparator works or not and the internal on-off state of the MOS tube array; the improved successive approximation logic control module has an estimation binary search mode, a reverse binary search mode, a bidirectional search mode and a freezing mode; the circuit has the advantages that the precision of output voltage is high, the circuit can be used for a high-precision on-chip power supply, the transient response speed is not limited by the scale of the MOS tube array, the transient response speed is high under any scale of the MOS tube array, obvious overshoot and undershoot phenomena cannot be caused, and the stability of the output voltage is high.

The invention discloses a large-scale MIMO beam training method and device, and the method comprises the steps: constructing a shared training system composed of a basic DFT codebook, a frequently repeated code target response matrix and a corresponding combined coding codebook, and carrying out the maximum posteriori decoding based on an energy observation sequence, thereby enabling the maximum posteriori decoding to be carried out under the time delay constraint condition that the number of training rounds is limited, and achieving the maximum posteriori decoding. And the beam forming vector with the maximum gain can still be accurately selected at a high probability. Compared with a traditional exhaustive search method, the inquiry frequency of beam training is remarkably reduced; compared with binary search and convolutional coding search schemes, the decoding accuracy is improved under the same number of training rounds, so that beam training performance improvement with low time delay and high accuracy considered is realized.

The invention discloses a Gossip protocol optimization method and system based on dynamic adjustment, and belongs to the technical field of distributed systems and network communication. The method specifically comprises the following steps: in a distributed system, acquiring and initializing network parameters; the method comprises the following steps of: executing previous R-1 rounds of propagation on nodes for updating messages of a distributed system according to a Gossip protocol, and constructing an expected number of updated nodes at the end of an ith round; calculating the total number of messages sent in the first R-1 rounds; constructing a CCP model, and solving the CCP model through a binary search method to obtain the minimum number of forwarding times; dynamically adjusting the last round of forwarding times K '; and executing the Rth round of propagation, forwarding messages to the K'random nodes by all updated nodes, and realizing Gossip protocol optimization based on dynamic adjustment. According to the method, through a staged propagation strategy and CCP modeling, the number of forwarding times K'of the last round is dynamically adjusted, and redundant messages are remarkably reduced.

The invention relates to the technical field of industrial automation, in particular to a self-adaptive baud rate negotiation method, system and device and a medium, and the method comprises the steps: if the device supports an extended baud rate, obtaining an available baud rate list commonly supported by a master station and the device; performing binary search circular screening according to the available baud rate list, and determining the highest available baud rate; performing a packet loss rate test according to the highest available Baud rate, and determining a working Baud rate; when the master station and the equipment stably work according to the working baud rate, checking the communication quality of the master station and the equipment according to a preset period to obtain an error rate; and if the error rates of a plurality of cycles continuously exceed the preset error rate threshold, selecting the Baud rate of the next gear of the working Baud rate from the available Baud rate list to execute the packet loss rate test to update the working Baud rate, thereby improving the data transmission performance under an excellent channel.

The present invention discloses a method and device for constructing an open circuit voltage-state of charge model of a battery, and relates to the technical field of battery management systems. When constructing an OCV-SOC model of a battery based on an SVR model, the present invention continuously performs binary segmentation and binary search on the initial search space of the parameters to be optimized in each dimension of the SVR model, so that the optimal search subspace can be quickly located from the large-scale initial search space of the parameters to be optimized, and then further performs detailed parameter optimization based on the optimal search subspace with a particle optimization algorithm, which greatly reduces the amount of calculation in the construction of the OCV-SOC model. At the same time, the binary segmentation and binary search of the initial search space can comprehensively evaluate the parameter performance of each search subspace in the large-scale initial search space, and can accurately determine the target search space of the PSO algorithm, avoiding the possibility of losing the optimal parameters, and improving the accuracy of the construction of the OCV-SOC model.

The present application relates to a kind of multi-unmanned aerial vehicle assisted wireless sensor network data acquisition method, belong to wireless sensor network technical field, comprising the following steps: S1: the maximum data acquisition time of UAV is obtained by dichotomy search method, and within the given time T, using time equal interval discrete technology is dispersed into a finite number of points with unmanned track;S2: according to initialization multi-UAV trajectory algorithm, solve UAV initial trajectory;S3: by the relaxation of binary variable, the original mixed integer non-convex problem is converted into continuous non-convex problem, and based on alternate optimization algorithm, the original problem is decomposed into two subproblems, by alternating iteration until convergence to obtain results, wherein, continuous convex approximation algorithm is used to solve subproblem, obtain UAV trajectory, uplink energy transmission scheduling and downlink data transmission scheduling;Until dichotomy search method result converges, the maximum time of UAV data acquisition is obtained.

This invention provides a microservice optimization method for industrial edge networks based on dynamic orchestration and migration expansion, comprising: abstracting the industrial edge network into an intelligent industrial network model composed of heterogeneous edge nodes and links; modeling industrial application requests as request chains composed of multiple microservices with data dependencies; calculating microservice instance deployment schemes and request routing schemes; statistically analyzing time-varying request flows; and utilizing a method based on... m / m / c This invention utilizes a multivariate performance analysis model of queue networks and a binary search algorithm to analyze and statistically determine the minimum number of cores that satisfy latency constraints. It then dynamically scales and adjusts the core resources of deployed microservice instances. Based on the preliminary dynamic scaling and adjustment results and migration costs, it solves a mixed-integer nonlinear programming migration decision problem to optimize the optimal collaborative deployment and request routing of microservice instances among edge nodes. This invention improves the service continuity and overall energy efficiency of industrial edge network systems.

The application provides a method and system for improving the diversity of bit-level autoregressive image generation, and relates to the technical fields of artificial intelligence and computer vision, and has the characteristics that in the sampling process of a bit-level autoregressive image generation model, for each generation scale k, the original prediction logits of all token bits at the scale are obtained; based on the original prediction logits, an adaptive temperature coefficient τk is dynamically solved by a binary search algorithm, so that after the logits are scaled by the τk and normalized by softmax, the difference between the actual average maximum bit probability of the current scale k and the preset target average maximum bit probability Sk is less than the preset tolerance; an adaptive temperature coefficient τk is dynamically determined by binary search, so that after the softmax normalization is applied to the τk. The application has the advantages that the generation diversity is improved by adaptively smoothing the early prediction distribution, and the visual quality is maintained by means of energy-based path search; the method mainly operates in the early stage with low computational overhead, and the inference delay is only slightly increased, and the core design can be extended to similar autoregressive generation models.

The invention discloses an over-path traffic solving method for a completely-crowded large-scale public transportation network, and the method comprises the steps: obtaining a getting-on arc and a corresponding effective departure frequency through a first algorithm; the second algorithm calculates the use probability of passing through all the upper turning arcs under the super path, so that the complete upper turning arc flow distribution of the super path, the use probability vector of the upper turning arcs and the use probability vector upper turning arcs of the nodes are obtained; then, adopting column generation to construct an OD pair sub-problem, and determining the optimal transfer amount from a non-reference super path to a reference super path in the OD pair sub-problem through binary search; and the fourth algorithm obtains an over-path traffic equilibrium solution under the public transportation network equilibrium condition. Therefore, by setting the first algorithm, the second algorithm and the like, a complete algorithm capable of processing large-scale public transportation network distribution is constructed, so that the super path is constructed based on the effective departure frequency, and the method is suitable for practical application of the city-level large-scale public transportation network.

The system comprises an identification module and a comparison module, the identification module is used for new energy controller electromechanical model parameter identification, the parameter identification is based on hardware-in-the-loop measured data, and the comparison module is used for comparing the hardware-in-the-loop measured data with the hardware-in-the-loop measured data. The value of the drop impedance in the identification model is automatically modified by writing a binary search optimization algorithm until the drop voltage identification precision requirement is met, and an optimal identification result is automatically output; the comparison module is used for performing new energy electromechanical transient modeling test, verifying the correctness of a drop impedance identification result in the identification module, automatically performing electromechanical transient simulation by setting different test working conditions, and automatically performing error analysis on each simulation result and a hardware-in-loop test result; according to the method, traditional manual testing work is replaced, parameter identification and electromechanical transient modeling testing can be automatically carried out, result data are automatically stored, and the precision and collection efficiency of a new energy electromechanical simulation model are greatly improved.

The present invention discloses a recomputation-aware model splitting method and system for pipeline parallel training, belonging to the field of artificial intelligence in computer science. The method of the present invention specifically includes: using the Dijkstra algorithm and two-layer binary search to calculate the minimum value of the sum of the forward propagation and backward propagation times of the pipeline microbatch under recomputation-aware model splitting; and using the intermediate results obtained by the Dijkstra algorithm during the search process to calculate and output a recomputation-aware approximately optimal model splitting strategy. The present invention utilizes the model characteristics of the Transformer large model and the characteristics of the pipeline parallel mode, designs a model splitting method and system, non-uniformly splits the model into each stage of the pipeline, and ensures the computational load balance of each pipeline stage when recomputation is enabled, thereby improving the overall performance of large model pipeline parallel training and the utilization efficiency of cluster devices.

An apparatus, system, and method for voltage regulator (VR) control are provided. An apparatus can include first, second, and third comparators configured to determine whether a load voltage (VLOAD) drops below a lower non-linear control (NLC) threshold, drops below a lower linear control (LC) threshold, and exceeds an upper LC threshold, respectively. The apparatus can include power gates (PGs) configured to adjust an output voltage (VOUT) based on a provided power gate (PG) code. The apparatus can include voltage regulator (VR) controller circuitry comprising synchronous LC circuitry and asynchronous NLC circuitry, the LC circuitry configured to increment or decrement the PG code responsive to the VLOAD dropping below the LC threshold and exceeding the upper LC threshold, respectively, and the NLC circuitry configured to increase the PG code based on a number of consecutive NLC droop events and responsive to the VLOAD dropping below the lower NLC threshold.

The invention relates to the technical field of cross-border e-commerce logistics order management, in particular to a cross-border e-commerce logistics order management method and system based on big data. Firstly, a current cross-border e-commerce order and current order information are acquired; then, inputting the order product information in the effective current cross-border e-commerce order into an order importance evaluation formula to obtain the importance degree of the current cross-border e-commerce order; order classification management is carried out according to the importance degree of the current cross-border e-commerce order; then, the historical residence time, the historical circulation expenditure and the historical product loss rate of each logistics node are evaluated, and an order logistics risk evaluation value is obtained; carrying out transportation risk management on each logistics node according to the order logistics risk assessment value; and finally, inputting the screened historical order logistics information and the order logistics risk assessment value into a multi-target logistics path optimization model, and performing optimization by using a binary search-minimum cost flow algorithm to obtain an optimal logistics transportation path of the current cross-border e-commerce order.