Abstract: Methods and systems for generating an integer neural network are described. The method includes receiving an input vector comprising a plurality of input values. The plurality of input values are represented using a desired number bits. The input vector is multiplied by a weight vector, and the products of which are summed to obtain a first value. The first value is quantized and applied to a piecewise linear activation function to obtain a second value. The piecewise linear activation function is a set of linear function that collectively approximate a nonlinear activation function. The second value is quantized to generate the output of the neuron in the integer neural network.

Abstract: This application discloses a notification display method. The method comprises: when there is a to-be-notified event, displaying, by a terminal, a notification by classification based on a notification type of the notification corresponding to the to-be-notified event; and displaying, by the terminal, in different display modes based on a priority of the notification, wherein the different display modes comprise: displaying in expanded form or displaying in collapsed form in the corresponding display pane, or the different display modes comprise: displaying in a first mode or displaying in a second mode in the corresponding display pane, and content displayed in the first mode is more than content displayed in the second mode.

Abstract: This application relates to the field of artificial intelligence technologies, and discloses a machine learning model search method, a related apparatus, and a device. In the method, before model search and quantization, a plurality of single bit models are generated based on a to-be-quantized model, and evaluation parameters of layer structures in the plurality of single bit models are obtained. Further, after a candidate model selected from a candidate set is trained and tested, to obtain a target model, a quantization weight of each layer structure in the target model may be determined based on a network structure of the target model and evaluation parameters of all layer structures in the target model, a layer structure with a maximum quantization weight in the target model is quantized, and a model obtained through quantization is added to the candidate set.

Abstract: This application discloses a notification display method. The method comprises: when there is a to-be-notified event, displaying, by a terminal, a notification by classification based on a notification type of the notification corresponding to the to-be-notified event; and displaying, by the terminal, in different display modes based on a priority of the notification, wherein the different display modes comprise: displaying in expanded form or displaying in collapsed form in the corresponding display pane, or the different display modes comprise: displaying in a first mode or displaying in a second mode in the corresponding display pane, and content displayed in the first mode is more than content displayed in the second mode.

Abstract: Methods, systems, and techniques for utilizing image processing systems to measure damage to vehicles include utilizing an image processing system to generate a heat map of an image of a damaged vehicle, where the heat map is indicative of a damaged area of the vehicle, and determining at least one measurement of the damaged area based on the heat map and a depth of field indicator corresponding to the image. In some embodiments, the image processing system also determines one or more types of damage of the damaged area, and/or also generates a segmentation map of the depicted vehicle and utilizes the segmentation map in conjunction with the heat map to measure damaged areas and locations thereof on the vehicle depicted within the image. In some embodiments, the techniques include determining the depth of field indicator of the image or portions thereof.

Abstract: A tool used in an image processing system assists a user to train one or more statistical image models or classification engines (e.g., the CNN models) that are used to detect damaged areas on a vehicle, to detect damage types and/or to detect segments of the depiction of the vehicle. The tool enables a user to select and annotate various different training images to be used to train the models, wherein each of the training images depicts damage of one or more damage types to various different vehicles or automobiles (including automobiles of different years/makes/models).

Abstract: An image processing system analyzes each of a set of vehicle images to determine one or more types of vehicle damage present in the selected images. The image processing system uses a characterization engine in the form of a neural network based image model to process each of the pixels of each of the selected images to determine, for the particular pixels of the image (or of the object depicted within the image) at which damage exists, the type of damage that is present. The damage types may include a tear, a scratch, a dent, a missing piece, a misalignment or other types of damage.

Abstract: A recommendation method applied to a terminal includes a rule platform with system-level permission, the rule platform is configured to save system resources and remind a user of a function of an application. The rule platform obtains a first event and a first rule related to the first event from a rule base. The rule platform includes the rule base including a rule and each rule includes a condition. The rule platform updates a matching status of a first condition corresponding to the first event in the first rule. The terminal displays a recommendation result corresponding to the first rule when a matching status of each condition in the first rule indicates that the matching is successful. The first rule is one of the rule related to the first event.

Abstract: A neural network model processing method includes obtaining a first low-bit neural network model through training, where the model includes a first operation layer and a second operation layer. Each operation layer includes at least one operation. Values/a value of a parameter and/or data used for the operation are/is represented by using N bits, and N is a positive integer less than 8. The neural network model processing method further includes compressing the model to obtain a second low-bit neural network model, where the compressed model includes a third operation layer. The third operation layer is equivalent to the first operation layer and the second operation layer, and an operation layer other than the third operation layer in the at least one operation layer is the same as an operation layer other than the first operation layer and the second operation layer in the at least two operation layers.

Abstract: A recommendation method applied to a terminal includes a rule platform with system-level permission, the rule platform is configured to save system resources and remind a user of a function of an application. The rule platform obtains a first event and a first rule related to the first event from a rule base. The rule platform includes the rule base including a rule and each rule includes a condition. The rule platform updates a matching status of a first condition corresponding to the first event in the first rule The terminal displays a recommendation result corresponding to the first rule when a matching status of each condition in the first rule indicates that the matching is successful. The first rule is one of the rule related to the first event.

Abstract: Embodiments of this application provide a recommendation method applied to a terminal, where the terminal includes a rule platform with system-level permission, and the rule platform is configured to save system resources and remind a user of a function of an application. The method includes: obtaining, by the rule platform, a first event; obtaining, by the rule platform, at least one rule related to the first event from a rule base, where the rule platform includes the rule base, the rule base includes at least one rule, each rule includes at least one condition; updating, by the rule platform, a matching status of a first condition corresponding to the first event in the at least one rule related to the first event; and when a matching status of each condition in a first rule indicates that matching is successful, displaying, by the terminal, a recommendation result corresponding to the first rule, where the first rule is one of the at least one rule related to the first event.

Abstract: This application discloses a message processing method and an apparatus. The method in this application includes: in response to a first operation performed by a user on a terminal, displaying a first unread message and a second unread message on a first user interface of the terminal, where the first user interface is a drop-down message notification bar or a HiBoard; and after the first user interface is exited, in response to a second operation performed by the user on the terminal, if the first unread message and the second unread message are still unread messages, displaying the first unread message instead of the second unread message on the second user interface of the terminal, where the first unread message is a non-unimportant unread message, the second unread message is an unimportant unread message, and the second user interface is a drop-down message notification bar or a HiBoard.

Abstract: This application discloses a notification display method. The method comprises: when there is a to-be-notified event, displaying, by a terminal, a notification by classification based on a notification type of the notification corresponding to the to-be-notified event; and displaying, by the terminal, in different display modes based on a priority of the notification, wherein the different display modes comprise: displaying in expanded form or displaying in collapsed form in the corresponding display pane, or the different display modes comprise: displaying in a first mode or displaying in a second mode in the corresponding display pane, and content displayed in the first mode is more than content displayed in the second mode.

Abstract: A user profile generation method and a terminal. The terminal includes a user profile platform and a first application. The method includes: receiving, by the user profile platform, a first request sent by the first application, where the first request is used to request a current user profile; obtaining, by the user profile platform based on the first request, a first user profile type required by the first application; in response to the first request, obtaining, by the user profile platform according to a user profile model, a current user profile corresponding to the first user profile type, where the user profile model is included in the user profile platform; and sending, by the user profile platform, the obtained current user profile to the first application.

Abstract: The present disclosure relates to recommendation methods and terminals. One example method includes receiving, by a terminal, a service request of a recommendation application, where the service request includes a configuration file, the configuration file includes at least an algorithm parameter and portfolio information used to identify an algorithm portfolio structure, and the recommendation application is included in the terminal, invoking, by the terminal, at least one algorithm from an algorithm library of the terminal based on the algorithm parameter and the portfolio information, predicting, by the terminal and based on the at least one algorithm and user data of the terminal, at least one service currently required by a user of the terminal, where the user data of the terminal is stored in a personal database located in the terminal, and displaying, by the terminal, at least one icon corresponding to the at least one service.