Abstract: Software tables in the control plane of a network device are downloaded into the data plane hardware after a reboot of the control plane. Valid entries are identified and downloaded while reducing the downloading invalid or otherwise unused entries to improve download times and reduce disruption of traffic resulting from the download. In some instances, blocks of entries containing a high ratio of valid to invalid entries are identified and downloaded to the hardware. In some instances, entries in the software tables that differ from corresponding entries in the hardware (diff entries) are identified and downloaded. In some instances, chunks containing diff entries and non-diff entries are identified and downloaded.

Abstract: This application provides an image sensor (702) and image light sensing method. The image sensor (702) includes a red pixel (R), a green pixel (G), a blue pixel (B), and an invisible light pixel, where the red pixel (R), the green pixel (G), and the blue pixel (B) are large pixels, the invisible light pixel is a small pixel, and a light sensing area of the large pixel is greater than that of the small pixel. The red pixel (R), the green pixel (G), and the blue pixel (B) are arranged in a Bayer format. In this application, when color information is sufficient, light crosstalk caused by the small pixel to the large pixel can be reduced, and therefore a signal-to-noise ratio of the large pixel can be improved.

Abstract: The present disclosure provides a method and apparatus for detecting significance of promotional information, a device and a computer storage medium. The method comprises: extracting each frame of image in a video segment, taking the image as input of a significance detecting module, and obtaining an output result of the significance detecting model; based on the output result, determining a significance score of promotional information corresponding to said each frame of image; obtaining the significance of promotional information corresponding to the video segment, based on the significance score of promotional information corresponding to each frame of image.

Abstract: A method and apparatus for optimizing an Ad CTR estimation model are provided. The method includes: calculating a direction vector and a step vector based on data in a training set, wherein the direction vector and the step vector are associated with a first parameter vector, and the first parameter vector is a parameter vector of the Ad CTR prediction model; calculating an optimized first parameter vector by setting the first parameter vector, the direction vector and the step vector as inputs of an update function, and by using a second parameter vector, wherein the second parameter vector is a parameter vector of the update function; estimating an optimized second parameter vector according to an optimization target in a validation set, the optimization target is determined by using the optimized first parameter vector; updating the optimized first parameter vector by using the optimized second parameter vector.

Abstract: Implementations detailed herein include description of a computer-implemented method to migrate a machine learning model from one accelerator portion (such as a portion of a graphical processor unit (GPU)) to a different accelerator portion. In some instances, a state of the first accelerator portion is persisted, the second accelerator portion is configured, the first accelerator portion is then detached from a client application instance, and at least a portion of an inference request is performed using the loaded at least a portion of the machine learning model on the second accelerator portion that had been configured.

Abstract: The present disclosure provides a method and apparatus for detecting significance of promotional information, a device and a computer storage medium. The method comprises: extracting each frame of image in a video segment, taking the image as input of a significance detecting module, and obtaining an output result of the significance detecting model; based on the output result, determining a significance score of promotional information corresponding to said each frame of image; obtaining the significance of promotional information corresponding to the video segment, based on the significance score of promotional information corresponding to each frame of image.

Abstract: Techniques for organizing and presenting deals/commercial offers received by users in emails are provided. Emails directed to a user that contain commercial offers for the user are determined. The determined emails are stored in a deal folder for the user. A deal newsletter is generated that at least summarizes commercial offers contained in at least a portion of the emails stored in the deal folder. The deal folder may be displayed to show the user the received deal emails. Furthermore, the deal newsletter may be displayed to the user to summarize the received deals for the user.

Abstract: Systems and methods for placing ads in a block on a webpage are disclosed. Generally, two ranking models are trained using a first and second ads data set. The first model predicts a first click probability for each ad in the first ads data and rank the ads based on the eCPM. The second model is trained using the second ads data set comprising a subset of the first ads data set and interaction features related to ad position in the block. The second model predicts a second click probability for each ad in the second ads data set. An overall expected revenue for each arrangement of ads in the second ads data set is then calculated. The computer system selects the arrangement with maximum computed overall expected revenue and places the ads in the block on the webpage according to the selected arrangement.

Abstract: Techniques for organizing and presenting deals/commercial offers received by users in emails are provided. Emails directed to a user that contain commercial offers for the user are determined. The determined emails are stored in a deal folder for the user. A deal newsletter is generated that at least summarizes commercial offers contained in at least a portion of the emails stored in the deal folder. The deal folder may be displayed to show the user the received deal emails. Furthermore, the deal newsletter may be displayed to the user to summarize the received deals for the user.

Abstract: Systems and methods for placing ads in a block on a webpage are disclosed. Generally, two ranking models are trained using a first and second ads data set. The first model predicts a first click probability for each ad in the first ads data and rank the ads based on the eCPM. The second model is trained using the second ads data set comprising a subset of the first ads data set and interaction features related to ad position in the block. The second model predicts a second click probability for each ad in the second ads data set. An overall expected revenue for each arrangement of ads in the second ads data set is then calculated. The computer system selects the arrangement with maximum computed overall expected revenue and places the ads in the block on the webpage according to the selected arrangement.

Abstract: Systems and methods for placing ads in a block on a webpage are disclosed. Generally, two ranking models are trained using a first and second ads data set. The first model predicts a first click probability for each ad in the first ads data and rank the ads based on the eCPM. The second model is trained using the second ads data set comprising a subset of the first ads data set and interaction features related to ad position in the block. The second model predicts a second click probability for each ad in the second ads data set. An overall expected revenue for each arrangement of ads in the second ads data set is then calculated. The computer system selects the arrangement with maximum computed overall expected revenue and places the ads in the block on the webpage according to the selected arrangement.

Abstract: Systems and methods for placing ads in a block on a webpage are disclosed. Generally, two ranking models are trained using a first and second ads data set. The first model predicts a first click probability for each ad in the first ads data and rank the ads based on the eCPM. The second model is trained using the second ads data set comprising a subset of the first ads data set and interaction features related to ad position in the block. The second model predicts a second click probability for each ad in the second ads data set. An overall expected revenue for each arrangement of ads in the second ads data set is then calculated. The computer system selects the arrangement with maximum computed overall expected revenue and places the ads in the block on the webpage according to the selected arrangement.