3results about How to "Reduce the error value" patented technology

The present disclosure provides a neural network-based weight quantization method, device and electronic equipment. The method comprises: obtaining calibration data corresponding to a calibration data set of the neural network; obtaining first quantized weights corresponding to a preset first truncation ratio of original weights of the neural network, determining a first quantization error of the first quantized weights relative to the original weights according to the calibration data; in the case where the first quantization error is determined to satisfy a preset convergence condition according to the calibration data, performing truncation processing and quantization processing on the original weights of the neural network according to the first truncation ratio to obtain network weights of the neural network. This method can automatically determine an adjusted first truncation ratio, so that the error value of the network weights of the neural network is smaller.

This application proposes a stepper motor drive correction system and method, relating to the field of stepper motor drive technology. The system includes an H-bridge, a sensing circuit, a sampling circuit, a current-to-voltage conversion circuit, a sine wave generator, a proportional amplifier, a comparator circuit, and a calibration control circuit. The sensing circuit samples the current of the low-side switch transistor of the H-bridge, which is then converted into a sensing voltage by the current-to-voltage conversion circuit according to a first preset ratio. The sine wave generator produces an ideal sine waveform, which is scaled by the proportional amplifier according to a second preset ratio to form a reference voltage. The comparator circuit compares the sensing voltage with the reference voltage and outputs the comparison result. Based on the comparison result, the calibration control circuit calibrates the first and second preset ratios sequentially, gradually reducing the error between the two voltages. Through these multiple calibrations, the system error is reduced to below a preset threshold, significantly improving the driving accuracy and smoothness of the stepper motor in high-microstepping mode.

This utility model relates to the field of separation membranes, and particularly to a sample testing device for separation membrane research and development. It includes a testing chamber with an insertion interface at the center of its top outer wall. A fixing structure is installed at the insertion interface, and partition plates are installed on the inner wall of the testing chamber at both ends of the fixing structure. The fixing structure includes a fixing frame, a placement groove on one side of the outer wall of the fixing frame, a pressing frame slidably connected in the placement groove, fastening grooves equidistantly located on one side of the outer wall of the fixing frame, and fastening protrusions equidistantly welded to the outer wall of the pressing frame at the fastening grooves. In this utility model, the detection gas enters the testing chamber under negative pressure. After passing through the separation membrane, the detection gas enters the other end of the testing chamber. An air quality sensor detects the air quality inside both ends of the testing chamber, facilitating detection, enabling quantitative detection, improving detection accuracy, and reducing error values.