115results about How to "Reduce learning" patented technology

The invention discloses a method for selecting a preset value of an image-guided ultrasonic diagnostic apparatus. Information about general information of the preset value, a source preset value and modification parameters is automatically and associatively stored by fully utilizing important information about use habits of doctors, historical check images of patients and the like when the doctors store an image; and the information is visually displayed to the doctors through various user interfaces so as to rapidly acquire the preset value expected by the doctors under the guidance of the image information. The method greatly reduces the difficulty of adjusting the parameters of the preset value and the doorsill for the use of the ultrasonic diagnostic apparatus, reduces workloads of the doctors for manually adjusting the parameters, and enables less experienced doctors to rapidly adjust the optimized image quality suitable for the patients and visually learn the parameter adjusting experience from images adjusted by the experienced doctors so as to shorten own learning process.

The invention relates to an electromyographic signal classification method based on a multi-kernel support vector machine. For a sample with complex distribution, based on the classification performance of a single-kernel support vector machine, the classification accuracy and the quantity of support vectors are easily influenced. The method combines a multi-kernel support vector machine method with a binary tree combination strategy and comprises the following specific steps of: collecting electromyographic signals of the lower limbs of a human body through an electromyographic signal acquisition instrument; denoising the electromyographic signals containing interference noise by using a wavelet coefficient inter-scale correlation denoising method; extracting the features of the denoised electromyographic signals to obtain the features of the electromyographic signals by using denoised wavelet coefficients; and classifying on the basis of the multi-kernel support vector machine. The method can well meet the multi-classification requirement of lower extremity prosthesis control, and takes into account both accuracy and instantaneity, and has broad application prospects in the multi-movement mode recognition of intelligent prosthesis control.

Disclosed is a communication network management system for converting action parameters contained in abstract requirements (abstract policy information) regarding a network to parameters conforming to the network technology (ATM, SDH, WDM, etc.) and type of network element to be set, and setting these parameters in the element. Specifically, a policy administration portion converts action parameters contained in abstract policy information to parameters dependent upon network technology and a policy enforcement portion converts the parameters obtained by this conversion to parameters dependent upon type of network element to be set and sets these parameters in the element.

The invention discloses a company announcement processing method for multi-task learning and a server, and the method comprises the steps: inputting historical announcement data into a sharing layer of a multi-task learning model, and carrying out the pre-training of the historical announcement data through Bert; inputting the data set corresponding to the processing task into a task layer of themulti-task learning model to train the multi-task learning model; obtaining current announcement data, and inputting the current announcement data into the trained multi-task learning model to obtaina task processing result. The multi-task learning model is constructed through using in a transfer learning and multi-task learning mode, and the method has the advantages of being high in learning efficiency, high in generalization, low in manual maintenance cost, high in accuracy of multiple tasks, high in recall rate and convenient for engineering deployment and maintenance.

The method includes steps: network device checks-up number of port in underling virtual local area network (VLAN); when number of port in VLAN is less or equal to 2, not learning media access (MAC) control address, broadcasting received data message in VLAN. The invention can reduce learning MAC address, raises user capacity of device, operating speed of network and extended performance.

The invention discloses a voiceprint recognition method based on gender, nationality and emotion information. The method specifically comprises the following steps: firstly, data preprocessing, secondly, feature extraction, and secondly, neural network parameter training: according to the specific structure of a neural network, in the training process, the input sequence of training sentences is disrupted, then 128 sentences are randomly selected as a training batch, and the number of data iterations is 80; wherein a training file required by the scoring fusion tool is a development set resultand a test set result of each system; wherein VOXCELEB1 test is used in the test set; wherein the development set is a test file which is generated based on 1211 training speaker statements and comprises 40,000 test pairs; and a final test set scoring result is obtained through 100 times of iteration. According to the invention, the recognition rate is improved.

The invention discloses a control method capable of intelligently selecting drying modes, an intelligent clothes drying machine and a storage medium. A plurality of drying modes set by a user are received and stored, and connection with a client is established; an drying mode selection command sent by the client is detected, and the intelligent clothes drying machine is controlled to work in the drying mode corresponding to the drying mode selection command; a receiving cloud platform recommends the drying mode with the largest matching rate according to the information of the environmental where the intelligent clothes drying machine is located, the information of the dried clothes and the drying habit, and the intelligent clothes drying machine is controlled to perform corresponding drying after entering the recommended drying mode. The intelligent clothes drying machine can greatly reduce the limitation of the drying environment, and perform clothes drying more conveniently and quickly through the selection and recommendation of the various drying modes.

The invention discloses an Android malicious software detection method based on a combined feature mode. Firstly, a certain amount of Android malicious software and Android benign software training samples are acquired to construct a training sample set; authority features and sensitive API features of each training sample are analyzed and combined to generate feature vectors of the training samples; the feature vectors of all the training samples serve as input to train an ELM, and the ELM is obtained; to-be-detected Android software serves as a test sample, and the authority feature and thesensitive API feature of the test sample are analyzed and combined to generate the feature vector of the test sample; the feature vector of the test sample is input into the ELM, and finally, whetherthe test sample is Android malicious software or not is judged by the ELM. The method has the advantages of being high in Android malicious software detection accuracy and short in learning time.

Provided in an embodiment of the present invention are a packet forwarding method and gateway, the method comprising: a first gateway receives a first Ethernet frame bearing an IP packet sent by a first virtual machine (VM), the Ethernet frame carrying a first source IP address field, a first source MAC address field, a first destination MAC address field and a first destination IP address field, the value of the first source IP address field being the IP address of the first VM, the value of the first source MAC address field being the MAC address of the first VM, the value of the first destination MAC address field being the identifier of a second gateway, and the value of the first destination IP address field being the IP address of a second VM; the first gateway changes the value of the first source MAC address field of the first Ethernet frame bearing an IP packet into an identifier of the first gateway so as to obtain a second Ethernet frame bearing an IP packet; the first gateway sends the second Ethernet frame bearing an IP packet to the second gateway. The present invention saves gateway resources and improves packet forwarding efficiency.

A system and method are provided for implementing a scheme to apply commercial web search technologies to biometric matching and identification based on converting biometric identification data to one or more text strings. Collected biometric identification information regarding particular physical traits is converted to a form that facilitates application of commercial Web search technologies to implement biometric matching and identification. A scalability of multi-modal biometric identification systems is maintained while substantially eliminating reliance on proprietary matchers and templates in support of interoperability and customer satisfaction. Separate biometric templates are converted into strings of searchable text in any combination of alpha-numerics during a standard biometric data enrollment process in order to limit the data storage requirements and streamline the later undertaken comparison process.

The invention relates to a #-shaped LBP and deep learning combination-based human face identification method, and belongs to the technical field of artificial intelligence. For the problems of insufficient feature extraction and classifier fitting of a conventional LBP, a human face identification method based on a #-shaped LBP and deep learning of a local texture feature is proposed. By utilizingan improved LBP algorithm, the local texture feature of a human face image is extracted, and an LBP histogram is established; then a deep belief network-based deep learning framework is established,the LBP histogram is input to a deep belief network, the network is trained by adopting an unsupervised layer-by-layer training method and a supervised BP algorithm, self-learning and self-optimization of the network are realized, and network parameters are obtained; and finally, the human face image is classified and identified by utilizing the DBN. According to the method, the distinctive texture feature of the human face image can be extracted; the superiority of the algorithm in identification rate is verified; and the robustness is relatively strong.

The invention discloses a sorting method for guiding a mechanical arm to grab materials with different poses based on a ConvPoint model, three-dimensional point cloud data of the materials is obtained through a 3D structured light camera, the three-dimensional point cloud data is combined with the ConvPoint model for material sorting, the method is not affected by the initial poses and appearances of the materials, various grabbing targets are achieved, and the system robustness is higher. Compared with an existing deep learning method, the method has the advantages that grabbing sampling is improved through PCA, the material grabbing accuracy is obviously improved, in addition, the point cloud in the network input clamp is evaluated instead of the point cloud of the whole object, the learning and reasoning time is shortened, and the material sorting efficiency is greatly improved. Besides, in the ConvPoint model training stage, the point cloud data of different types of complex structure materials are collected in real time to serve as a training set, the method does not depend on a data set of simple objects of an online open source, complex objects of different poses are grabbed, and the intelligent level and the automation level of a sorting system are improved.

The invention discloses a method and system for carrying out contextualization interactive processing. The method comprises the steps of acquiring state information of various system events on a mobile terminal, generating early warning data for each system event according to the state information, and carrying out prompting on a desktop of the mobile terminal when the phenomenon that the data value of the state information of one system event on the mobile terminal is lower than the early warning data of the system event is monitored. By means of the method and system for carrying out the contextualization interactive processing, an operation interface can be simplified, a uniformed interactive mode is used, the system events are monitored through background services, and necessary prompting information (such as low power, flux warning, high internal storage and other early warning prompting information of the system events) can be fed back and provided to users in time. A timely and accurate one-button type operating mode can be further achieved, optimization processing can be carried out on all the system events through the method and system, the operating experience of the users is simplified, more learning or adapting time is reduced, and meanwhile interactive communication with the users can be carried out in time.

The present invention provides compounds of formula I described herein. The present invention also provides a method of treating a cognitive dysfunction in a mammal. The method includes administering to the mammal an effective amount of a compound of formula I described herein (e.g., stearyl choline chloride).

The invention provides a reverse proxy method and device, electronic equipment and a storage medium. The reverse proxy method based on object storage comprises the following steps of receiving file request information, wherein the file request information is used for obtaining a file and at least comprises a filename and a file path; determining a data center where the file is located; obtaining auser name and a key, which are used for accessing the determined data center, from a configuration file, wherein the user name and the key for accessing one or more data centers are stored in the configuration file; encrypting the file request information, the user name and the key to generate a token; generating an access address for accessing the determined data center, wherein the access address comprises the token; and accessing the determined data center according to the access address in order to make the determined data center verify the token, and extracting the file from the determined data center. According to the method and the device which are provided by the invention, the file is obtained automatically.

The invention discloses a cloud computing-based pressure casting simulation control method and belongs to the technical field of pressure casting control. A pressure casting simulation control method in the prior art does not disclose a technical scheme about how to simulate a simulation model, and the prior art is narrow in application range, cannot be suitable for various pressure casting situations and is poor in simulation effect and high in usage cost. According to a cloud computing-based pressure casting simulation technology, a cloud computing mode is introduced in a pressure casting simulation technology to realize intelligent configuration of resources. Through a pressure casting cloud service platform, the learning and usage threshold of the pressure casting simulation control technology is greatly lowered. A pressure casting cloud solver is optimized according to the pressure casting simulation control method and then is packaged in a cloud service platform. The technical scheme of pressure casting simulation is detailed, feasible, wide in application range, accurate in computation, good in simulation effect, quick and convenient in use and low in usage cost.

The invention discloses a pre-fenestrated aortic covered stent and a delivery system and a utilization method thereof. The aortic covered stent with a large curved side and a small curved side comprises an outer covered stent and an inner covered stent, each of the outer covered stent and the inner covered stent is tubular and comprises a covered section and an uncovered bare stent crown, and thenear end of the covered section of the outer covered stent is provided with a rectangular uncovered pre-fenestrated hole at a reinforcement rib position of the large curved side. The inner covered stent internally sleeves the outer covered stent and identical to the outer covered stent in diameter, and the inner covered stent is shorter than the outer covered stent. The pre-fenestrated aortic covered stent and the delivery system and the utilization method thereof have advantages that due to positional adjustability of the pre-fenestrated hole in the aortic covered stent, fenestration operation is simplified, and fenestration success rate is increased.

The invention relates to a multi-scale target tracking method based on learning rate adjustment and multi-layer convolution feature fusion, and belongs to the technical field of computer vision tracking. The method comprises the following steps: firstly, extracting image features by adopting a layered convolutional neural network, and predicting a target position by fusing multi-layer convolutional features by utilizing a linear weighting method; determining the optimal scale of the target by using the target convolution features under multiple scales; and finally, evaluating the confidence coefficient of target response by utilizing average peak value related energy, evaluating the motion condition of the target according to the frame difference mean value and the displacement of two adjacent frames of target images, and adjusting the learning rate of the filter model according to the prediction position confidence coefficient and the appearance change of the target images. Accordingto the method, the traditional correlation filtering tracking method can be effectively processed; traditional manual features lack semantic information, targets and backgrounds cannot be effectivelydistinguished, error accumulation can be caused by indiscriminately updating a filter due to the fact that the learning rate is a fixed value under the conditions of shielding, target loss and the like, and the method can effectively track the targets under complex conditions.

The present invention relates to an implementation example of a method for adding a next MAC address on an FDB table, which comprises step 1, the adding request message of the next MAC address is transmitted; step 2, an FDB table in a hash unit is confirmed as an initial FDB table according to the initial index value; step 3, whether the FDB table is available or not is detected, and if the FDB table is available, the next MAC address is added to the FDB table; if the FDB table is not available, an FDB table in another hash unit is confirmed as the next FDB table, and the step is repeated till an available FDB table is detected, and added with the next MAC address. The present invention relates to an implementation example of a FDB equipment, which comprises FDB table directional units connected in sequence, an FDB table state detection unit and an adding unit of the next MAC address. In the implementation example of the present invention, the next MAC address is uniformly added to the FDB table, thus the resource utilization rate of the FDB Table is improved, and the learning of a two-layer forwarding address and the impaction of the adding are reduced.

The invention is based, in part, on the discovery that secretin and secretin analogs can be used to treat affective disorders that include symptoms of anxiety. Secretin and secretin analogs can be administered to patients to treat affective disorders, such as mood, psychotic, mental, substance-abuse withdrawal, eating, and sexual disorders, as well as other disorders with attendant sensory processing dysfunction, including symptoms of anxiety and / or fear. In addition, secretin and secretin analogs can be administered to improve cognition, including learning and memory, as well as to lessen the effects of conditioned startle. Secretin has the added benefit of causing no negative side effects upon administration and being easy to administer.

The invention discloses a training method of a self-adaptive conditional random field algorithm for automatically news splitting. The method comprises the following steps of: 1, digitizing a news program video, and extracting news splitting feature data and a news splitting label according to the news program video; 2, defining a fixed learning template, and recording learning step lengths including a feature step length and a label step length in the learning template; 3, adaptively adjusting the step length by adopting a heuristic method according to the fixed learning template and the training data news splitting label; and step 4, learning parameters in the conditional random field algorithm by adopting a gradient descent method according to the data learned by the adaptive method in the step 3. According to the method, the current news story feature data and a plurality of previous news story feature data and next news story characteristic data are learned in a self-adaptive manner according to the conditions of the training data, so that learning of the current news story characteristic data and the adjacent news story characteristic data is more focused, learning of non-adjacent news story characteristic data is reduced, and the method has great significance in improving the news automatic splitting accuracy of conditional random field algorithms.

The invention discloses an anti-daze bracelet. The bracelet comprises a vibration sensor module, a control module and a vibration module; the vibration sensor module and the vibration module are respectively and electrically connected with the control module; the control module is used for receiving a vibration sensor signal and controlling the vibration module to vibrate. The above structure is capable of monitoring whether the user is in a daze state in real time, the judgment of the daze state is triggered through the measurement of the vibration and the setting of the vibration interval, and the bracelet can vibrate after monitoring the daze state, thereby reminding the user of leaving the daze state. Furthermore, the device is set as the bracelet state to achieve the decoration effect; and the bracelet is capable of directly stimulating the skin, and effectively guaranteeing the timing detection and timing reminding of the daze state. Furthermore, the device is convenient for use, the daze state can be detected only needing to wear the bracket on the hand, and the usability of the device is improved.

The invention discloses a method for controlling a single-lens reflex camera to carry out panoramic shooting through a mobile phone. The method comprises the steps that the smart mobile phone is placed on a single-lens reflex camera clamping base, previewing of the single-lens reflex camera and a mobile phone assistant APP are started, the single-lens reflex camera is moved to select a scenery spot to be shot, the mobile phone displays the scenery spot selected by the single-lens reflex camera, the assistant APP prompts and guides a user to move the single-lens reflex camera to an optimal shooting position, the assistant APP sends a shooting instruction to the single-lens reflex camera, the single-lens reflex camera shoots a single picture, and the assistant APP obtains the shot picture from the single-lens reflex camera and sends the picture to a storage card of the mobile phone; if the number of pictures reaches a preset number or the user stops shooting the pictures, the shot pictures are composited to be a panoramic picture, and shooting is ended; or, other scenery spots are selected for shooting. The optimal spot for shooting the single picture in a panoramic mode through the single-lens reflex camera is determined and the panoramic picture is composited through the smart mobile phone, shooting quality is improved and real-time operability is high.

The invention is suitable for the technical field of touch screens and particularly relates to a method and a system for regulating handwriting recognition speed and touch screen equipment. The method comprises the following steps of: recording an instant speed for handwritten input of a user and a time interval for inputting two adjacent characters; calculating an average speed and an average time interval for the handwritten input of the user according to the instant speed and time interval; and regulating the handwriting recognition speed into the average speed and the average time interval. In the embodiment of the invention, the handwriting recognition speed can be automatically regulated. In addition, the method, the system and the touch screen equipment ensure that the user can more accurately recognize the characters input by the handwriting of the user from new equipment trial to more skillful use, make the operation of a mobile phone terminal more humanized and intelligent, reduce the learning burden and adapting burden of the user, do not need debugging and have friendly interactive experiences.

The invention belongs to the technical field of artificial intelligence, and particularly relates to a system suitable for driverless decision learning and training. As an algorithm framework is designed in an existing method, and big data is used for training, or training / verification is carried out according to a set of formulated rules, the efficiency is low, and the cost is high. The inventionprovides a system suitable for driverless decision learning and training. The system comprises a virtual reality / machine learning unit, wherein the virtual reality / machine learning unit comprises a main processor module; the main processor module is connected with a virtual reality presentation module; the main processor module is connected with a vehicle control electromechanical module; the vehicle control electromechanical module is connected with a steering wheel; the vehicle control electromechanical module is connected with gears; and the vehicle control electromechanical module is connected with an accelerator / brake. The problem of how to quickly train an algorithm in the unmanned driving technology is solved, and the vehicle driving ability, especially artificial intelligence learning, is improved.

The invention relates to the field of battery swap service and specifically relates to a charging and battery swap station monitoring system based on B / S architecture and the charging and battery swap station. The invention aims at solving the problem that according to an existing human-machine interaction system, a software platform is limited by an operating system, the updating and iteration are slow and the development cost is high. In order to achieve the purpose, the charging and battery swap station monitoring system provided by the invention comprises a server side and at least one HMI (Human Machine Interface) side. The server side can obtain real-time operation data of the charging and battery swap station and the server side can convert the real-time operation data into a form capable of being directly accessed by the HMI side. The HMI side can directly access the real-time operation data and the HMI side can monitor and control a charging and battery swap process based on the real-time operation data. The development of the HMI side is not limited by the software platform in the setting mode, the updating and iteration are relatively convenient and fast, the configuration flexibility of the HMI side is increased, and the development cost of the HMI side is reduced.

The invention relates to a local accent speech recognition system based on an embedded mobile device. The system comprises a model training module, a feature extraction module and a pattern matching module which are integrated on an embedded mobile device. The model training module is used for collecting and training local accent speech to get an entry model of local accent. The feature extraction module is used for extracting the speech features of an input local accent. The pattern matching module is used for carrying out speech matching calculation on the speech features according to the entry model to get a speech recognition result. The local accent speech recognition system based on an embedded mobile device of the invention can improve the recognition rate of dialectal speech, reduce the difficulty of learning and training during speech recognition used by dialect speakers, and reduce the workload of learning and training. Meanwhile, the system introduces the speech recognition technology to a variety of embedded mobile devices, and can realize intelligent interaction.