9 results about "Window selection" patented technology

The application provides a risk prediction window selection method and device based on multi-target cooperative evaluation and electronic equipment, relates to the technical field of machine learning and data modeling, and comprises the following steps: obtaining time sequence historical data of a target object, and dividing the time sequence historical data into a training subset and a test subset used for simulating future prediction according to time sequence; a candidate prediction window set containing at least two different time lengths is set, for each window length in the set, a corresponding binary classification prediction task label is defined based on the length, and a corresponding prediction model is trained by using the training subset; taking the model performance corresponding to the longest reference window in the candidate window as a reference, the optimal window length is selected from the candidate window set according to a preset selection criterion. The application ensures that the selected window length directly serves the accuracy target of the final prediction model, so that the shortest effective window is automatically found under the premise of ensuring that the prediction performance of the long window reference model is equivalent.

The invention discloses an adaptive multi-window and anti-noise fluctuating commodity sales volume prediction method, which comprises the following steps of: (1) constructing a multi-window network architecture, and carrying out parallel processing input through a plurality of sub-networks with different historical window lengths so as to explicitly model a long-short term multi-scale mode in a time sequence; (2) introducing an adaptive multi-window selection mechanism, and dynamically learning an optimal window length matched with the characteristics of the current input sample for each sub-network; and (3) designing a self-adaptive anti-noise module, and injecting anti-noise with variable intensity into the self-adaptive anti-noise module according to the window length of each sub-network so as to enhance the robustness of the model in a high-noise environment. According to the method, through an innovative multi-window network architecture, adaptive window selection and an adaptive anti-noise mechanism, the problems of multi-scale mode modeling and model robustness in a high-fluctuation and high-noise fluctuation commodity sales volume prediction task are effectively solved, and the model prediction accuracy and generalization ability are remarkably improved.

The invention belongs to the technical field of computer vision, and provides a single-eye emotion recognition sparse Transform modeling method based on event driving. The invention provides an event camera-oriented efficient visual Transform framework for eye emotion recognition, and a window selection and double-mask token selection combined pruning optimization mechanism is innovatively designed for event data characteristics and periorbital region analysis requirements. The method comprises the following steps: firstly, dynamically screening out a low-value region by quantitatively evaluating the information density of a local window so as to reduce calculation redundancy, and strengthening attention focusing on periocular key features; on this basis, a decision model fusing event stream time-space characteristics and token semantic relevance is further constructed, precise recognition and dynamic pruning of redundant tokens are realized by using binary masks, the calculation complexity is remarkably reduced on the premise of ensuring the classification precision, and an efficient solution is provided for real-time emotion recognition scenes with limited resources. The method provided by the invention can have remarkable advantages in a monocular emotion recognition task.

The application discloses a free window selection method applied to an array readout circuit, and a window selection circuit structure comprises a configuration module, an execution module and a transmission module. In the array configuration stage, the configuration module supports user-defined window size and position, the execution module activates the pixels in the window according to the configuration information, and forms a dedicated data transmission path; in the array quantization stage, the execution module selects a time-to-digital converter to quantize the photon time of flight; in the array transmission stage, the transmission module adaptively adjusts the transmission duration according to the window size, and sequentially outputs the quantized data from the dedicated data transmission path. The window selection circuit can provide various spatial resolution and frame frequency selection under the premise of ensuring consistent array time resolution, improve the application flexibility of the array, reduce the system power consumption, and has good portability.

The invention relates to a TSN off-line-on-line hybrid scheduling method based on an FPGA. The method comprises the following steps: an off-line stage: calculating a gating table of periodic flow by using SMT; in the mapping stage, the gating table is converted into a checkable two-dimensional Boolean table through serial analysis, address decoding and a dual-port RAM; in the on-line stage, the FPGA parallel LAW calculation and ASAP window selection assembly line dynamically distributes a CQF time slot for a burst frame in a microsecond level, and returns offset through a UDP (User Datagram Protocol). According to the invention, through the architecture of off-line calculation and on-line parallel table look-up, the coexistence of periodic traffic certainty guarantee and burst traffic microsecond scheduling is realized, and the real-time performance, the resource efficiency and the expandability are considered.

Methods and apparatus for retransmission using a window-based cross-block check block are described. An initial transmission of a first transport block having a first plurality of code blocks is transmitted. Feedback indicating that at least one code block is not successfully decoded is received, where the feedback includes an indication of an unsuccessfully decoded code block having a minimum index with respect to the first transport block. A retransmission of one or more cross-block check blocks is transmitted. The cross-block check block is generated according to a selected code block selected by a window, the starting point of which is defined according to the indication of the unsuccessfully decoded code block in the feedback.

The invention discloses a core unit, a high-performance frequency hopping carrier generator and a frequency hopping transmitter, and the core unit comprises a phase interpolator which is used for initializing a clock period signal with an adjustable phase, and the frequency of the clock period signal is the same as the frequency of a reference clock; the counter is used for outputting an integral multiple frequency hopping result of the reference clock; the delay adjusting module is used for adjusting the time sequence relation between the output of the phase interpolator and the output of the counter; the either-or selector is used for selecting the output of the counter; and the window selection module is used for selecting the required phase interpolator output according to the output of the counter.

The application discloses a kind of ground-based measurement and control resource scheduling method, system, equipment and medium, it is related to spaceflight measurement and control technical field, method includes: determining the visible window set VTM constructed by multiple to-be-scheduled spacecraft corresponding visible window, and multiple to-be-scheduled spacecraft corresponding task sequence R;In the visible window set VTM, the i th task in task sequence R is selected in window, and when all windows do not satisfy the constraint condition of i th task or when selected target window and i th task exist task arrangement conflict, update task conflict set Θ;Task conflict set Θ is handled by conflict back jump method, determine the scheduling scheme corresponding to i th task.The application is matched in the initial visible window matching process without obtaining matching result by conflict back jump method, and then adjust matching search strategy for the second time visible window matching, improve the applicability of the scheduling of to-be-scheduled spacecraft.

A lightweight optimization method based on a DeepSeekR1 large model is characterized in that the dimension of a lexical vector is set to be 1024 at most, and only core semantics are reserved; setting a local window selection window, wherein each lexical element only focuses and pays attention to at most 256 adjacent lexical elements; grouping and solidifying expert routes according to fields; specialist ability is limited, and only universality, namely a high-frequency existing knowledge scheme, is reserved; phrases with complete meanings are directly generated according to the high-frequency associated words, only core terms are verified not to be subverted, that is, sentences and meanings are completely opposite, and only a universal scheme is reserved. Compared with the prior art, the method has the advantages that the computing power requirement during operation of a large model can be effectively reduced, the bandwidth requirement during internal data exchange is small, power consumption and heat emission are small, the energy consumption equipment cost is low, local deployment is facilitated, and the carbon reduction suggestion is changed from an expert report to a carry-on guide by sacrificing 30% of deep generalization ability, such as policy analysis, industrial emission reduction, exchange speed improvement and energy efficiency optimization.