6 results about "Forward recursion" patented technology

The invention provides a task termination strategy generation method and system based on a dynamic programming algorithm, a storage medium and electronic equipment, and relates to the field of risk control. The method comprises the following steps: firstly, defining that the deterioration state of a system obeys a multi-stage continuous time Markov chain and random task duration; secondly, constructing a value function to obtain expected costs of tasks in all possible execution durations in each deterioration state; and finally, a reverse induction method is combined with a dynamic programming algorithm, forward recursion is carried out from the end point time of the longest task duration, so that an optimal task termination decision under each time step length is determined, and global optimization can be realized. According to the method, the uncertainty of the task execution time in an actual system can be reflected more truly, the dynamic change in the task execution process can be monitored in real time, and the task termination strategy is dynamically adjusted based on real-time data, so that a complex and changeable task environment can be flexibly coped, and the survival rate of the system is maximized.

The invention discloses an energy storage optimal self-scheduling analytic solution method based on parameter dynamic programming, which comprises the following steps: defining decision variables, an objective function and constraint conditions of an energy storage optimal self-scheduling problem, and establishing the energy storage optimal self-scheduling problem; parameterization reconstruction and time period decomposition are carried out on the energy storage optimal self-scheduling problem, and a single-time-period energy storage optimal self-scheduling parameter planning sub-problem is established; gradually obtaining an optimal value function and an optimal solution function of each time period through a back-stepping recursion mode; and gradually obtaining an optimal scheduling decision of each time period through a forward recursion mode. The method does not depend on an external optimization solver, the global optimal solution of the problem can be obtained while the solving efficiency of the energy storage optimal self-scheduling problem is improved, and the method can be used for accelerated solving of various energy storage self-scheduling, electric vehicle cooperation and other problems.

The application discloses a kind of sea surface flow field inversion and reconstruction method based on stationary satellite high frequency image, belong to marine remote sensing technical field.It includes the following steps: satellite image is preprocessed and locally enhanced;Initial inversion is carried out using dense optical flow algorithm for single time phase flow field;The flow field data cube of space-time grid uniformization is constructed;Time series reconstruction is carried out using two-way Kalman filtering, through forward recursion, backward recursion and two-way fusion, using time series information to correct abnormal vector and fill in missing data;Finally, based on the physical consistency principle of seawater approximate incompressibility, the reconstructed flow field is subjected to iterative divergence constraint, and the non-physical disturbance is weakened.The application effectively solves the problem that the traditional optical remote sensing flow measurement method is easy to produce abnormal vector and poor spatiotemporal consistency under complex sea conditions such as cloud cover and weak texture by introducing time series correlation and physical constraint, significantly improves the accuracy, stability and physical reasonableness of sea surface flow field inversion.

The invention discloses an assembly line optimization method and system for accelerating multi-mode large model lightweight training, and the method comprises the steps: obtaining a data set, determining the batch size, setting the assembly line stage number, the micro-batch number and the required assembly line stage number according to the GPU number, and enabling the micro-batch number to be an integral multiple of the GPU number; dividing the to-be-trained model into parts with the number of assembly line stages according to the structure of the to-be-trained model, wherein the difference between the calculated amounts of different parts does not exceed 10%; deploying the models of different stages to a GPU according to a predetermined rule; and performing a training process on each GPU according to a pipeline parallel rule. According to the method, the characteristic that forward recursion and back propagation are close in time consumption in lightweight training of the multi-modal large model is utilized, bubbles in pipeline parallelization are reduced while low GPU video memory occupation is used, and the training efficiency is improved.

The application provides a CVT error calculation method based on an improved Kalman smoothing method and a medium, aiming to improve the error estimation accuracy and stability of the CVT system. The method combines specific CVT accuracy level calibration parameters, optimizes the real-time data of multiple in-phase CVTs through Kalman filtering and Kalman smoothing technology, and improves the measurement accuracy. In the method, the algorithm effectively reduces the influence of noise on CVT error calculation through iterative optimization of multi-time data, and improves the measurement accuracy and stability of the CVT system. Through the double optimization process of forward recursion and backward correction, the application realizes accurate correction of the CVT error in the dynamic changing working environment. The Kalman smoothing method can combine historical data and real-time data, not only improving the accuracy of single estimation, but also enhancing the stability and reliability of error calculation in multi-time data.

The invention discloses a novel Turbo code decoding state value normalization method, which comprises the following steps of: calculating N state values for a first bit of an input component decoder to form a state value sequence X; the state value sequence X calculated in the first step is compared with the maximum value in the positive direction and the minimum value in the negative direction one by one, the state values larger than the maximum value form a mark sequence Fmax, and the state values smaller than the minimum value form a mark sequence Fmin; judging the flag sequences Fmax and Fmin obtained in the step 2 to obtain a new state value sequence X '; taking the new state value sequence X'as input, and calculating N state values of the next bit; and repeating the step 2 to the step 4 to complete the calculation of the forward recursion values and the backward recursion values of all bits so as to complete the whole decoding process. According to the method, only fixed two-stage running water required by the state value needs to be calculated, and the time delay is greatly reduced.