9 results about "Ladder network" patented technology

The application relates to the technical field of elevator Internet of Things data processing, and discloses an elevator Internet of Things data management and backup method and system under a hybrid cloud architecture, which comprises the following steps: collecting elevator terminal equipment original data streams on the edge side and processing the original data streams to generate pretreatment data; identifying data sensitivity according to a security rule library, encrypting high and medium sensitive data, and keeping low sensitive data in plaintext; dividing the data into three backup priority sets according to business criticality, and respectively distributing the data to an edge side local backup unit, a private cloud and a public cloud backup cluster after clustering based on access frequency and storage occupation; monitoring storage load, link state and backup progress, and scheduling resources when an exception occurs; and regularly recovering and practicing the backup data. Through data hierarchical processing, differential storage, dynamic monitoring and scheduling and regular practice, the application supports intelligent management of large-scale elevator cluster data.

The invention is suitable for the technical field of model fine tuning, and provides a model fine tuning method and device, electronic equipment and a storage medium, and the method comprises the steps: determining a to-be-adjusted network layer from a ladder side network of a to-be-fine-tuned model; a new rank corresponding to the network layer to be adjusted is determined according to the rank of the original parameter matrix of the network layer to be adjusted, a target rank is obtained, and the target rank is smaller than the rank of the original parameter matrix; compressing an original parameter matrix of the network layer to be adjusted into at least two low-rank parameter matrixes based on the target rank, wherein the rank of the low-rank parameter matrixes is the target rank; determining the adjusted ladder side network according to the at least two low-rank parameter matrixes; and training the to-be-fine-tuned model by adopting a training data set, and updating the at least two low-rank parameter matrixes in the adjusted ladder side network according to a training result to obtain the fine-tuned to-be-fine-tuned model. The training efficiency of model fine tuning can be improved.

The invention discloses an adaptive method during open world testing based on a hierarchical ladder network. The method comprises the following steps: inputting each target domain image into a pre-trained source domain model to obtain a first category prediction result and an entropy value; determining a sample type of each target domain image in the current batch according to the entropy value; if the sample type of the target domain image is an out-of-distribution sample, generating a second category prediction result based on a hierarchical ladder network set in the source domain model and the category mark of each layer of encoder; fusing the first category prediction result and the second category prediction result to obtain an out-of-distribution discrimination score; network parameters of the hierarchical ladder network are updated through back propagation, and a target domain model is obtained; and determining the sample type of the input current target domain image based on the hierarchical ladder network in the target domain model, and outputting a classification result. The problem that the prediction accuracy of the adjusted model is low due to the fact that the model is prone to misleading after the samples outside the distribution and the samples inside the distribution are mixed is solved.

The invention relates to the technical field of elevator internet-of-things data processing, and discloses an elevator internet-of-things data management and backup method and system under a hybrid cloud architecture, and the method comprises the steps: collecting an original data stream of elevator terminal equipment at an edge side, and carrying out the processing of the original data stream to generate preprocessing data; identifying data sensitivity according to a security rule base, encrypting high and medium sensitivity data, and reserving plaintexts for low sensitivity data; dividing into three-level backup priority sets according to business criticality, and after clustering based on access frequency and storage occupation, respectively distributing to an edge side local backup unit, a private cloud backup cluster and a public cloud backup cluster; the storage load, the link state and the backup progress are monitored, and resources are scheduled when abnormity occurs; and performing recovery drilling on the backup data regularly. Intelligent management of large-scale elevator cluster data is supported through data grading processing, differentiated storage, dynamic monitoring scheduling and regular drilling.

The present invention relates to a circuit, a method, and a system for balancing the voltage of storage units (BT) connected in series. The circuit corresponds to a ladder network in which horizontal switches (SH) form the rungs of the ladder, a storage unit (BT) is arranged between each rung on the first ladder, and a vertical switch (SV) is arranged between each rung on the second ladder. By switching the horizontal switches (SH) and vertical switches (SV), individual storage units (BT) can be connected to a balancing source (SY), which adjusts the voltage of the connected storage unit (BT), thereby balancing the voltage of the storage units (BT) in the series connection.

The invention relates to the technical field of data transmission, and discloses a management platform for optimizing elevator networking equipment identity and data by using HyperledgerFabric, comprising an elevator abnormity identification unit which distributes a unique identity label for each elevator equipment, collects elevator equipment operation data, identifies abnormal data and obtains a final risk prediction value through a HyperledgerFabric block chain network; the maintenance request generation unit is used for collecting maintenance request book templates, calculating template characteristic values according to keywords, dividing template grades, determining template grades corresponding to the maintenance request book templates, determining the maintenance request book templates according to the matching degree of each template, and generating a maintenance request book; and the maintenance request sending unit is used for determining maintenance personnel and sending the maintenance request book to the maintenance personnel based on the block chain network. According to the invention, credible distribution is carried out based on the block chain network, and the transparency, timeliness and non-repudiation of maintenance task distribution are ensured.

A waveguide combiner is provided. The waveguide combiner includes a waveguide combiner substrate. The waveguide combiner provides a plurality of staircase structures arranged on the substrate. Each staircase structure includes a plurality of steps. Each step has the same step width. The plurality of steps has a trim width from the first step to the last step. The plurality of steps includes an upper step having an upper width.

The application discloses a resistance ladder network-based electric stimulation auxiliary positioning wristband and a regulating method. The wristband comprises a host computer, a functional wristband and an electrode module. The functional wristband is internally embedded with mutually insulated conductive conductors. The functional wristband internally embeds the conductive conductors, and is provided with through holes. Adjacent through holes are connected in series with resistors to form a resistance ladder network, so that each through hole is an internal resistance increment access node. The electrode module comprises electrode pieces and conductive columns, which are detachably fixed to the access nodes and electrically connected with the conductors through installation channels. The host computer detects internal resistance increments to identify electrode positions. The method is as follows: impedance decoupling identifies positions, deducts line resistance to obtain net impedance; impedance characteristics identify acupoints; when impedance falls into a nerve interval and presents a minimum value, it is determined that the acupoint is aligned; anatomical characteristics are inferred according to position numbers to call stimulation parameters; skin electric response is closed-loop regulated; charges are discharged before signals are collected in pulse intervals; and parameters are adjusted according to skin electric slopes. The application constructs an intelligent system, and improves electric stimulation accuracy and user experience.

The application claims a segmented R-2R inverting ladder resistor network with high spurious-free dynamic range, which improves the spurious-free dynamic range (SFDR) of the circuit by segmenting the R-2R inverting ladder resistor network and performing non-overlapping rotation selection processing on the high-bit segment resistor network. The circuit mainly comprises: a thermometer decoder for converting a high-4-bit binary code into a 15-level thermometer code; an accumulator for accumulating the high-4-bit binary code to generate a shift control signal (also called a pointer) required by a logarithmic shifter; a logarithmic shifter for performing a shift operation according to the shift control signal input by the accumulator and the thermometer code input; a level holding circuit for compensating for the level lost during the shift operation of the logarithmic shifter; a latch for time-domain alignment of the signals of the low-12-bit and high-4-bit; and a segmented R-2R inverting ladder resistor network for receiving the digital signal of the latch and then decoding the digital signal.