Lemon

A machine learning model receives sequences of time series data representing a plurality of instances associated with an event in a cloud network. Each sequence of time series data can be projected into a high-dimensional space. Positional encoding is performed to add positional information to each sequence of time series data. Spatial attention weights are distributed across different timesteps in each sequence of time series data by the spatial attention layer to capture the local temporal structure of the sequence data. Context information encoding is performed on an output from the spatial attention layer by transformer encoder layers of the machine learning model. A representation of each sequence of time series data is generated based on outputs from the transformer encoder layers for predicting a probability that each instance is impacted by the event.

A method, device and storage medium for content delivery are provided. In the method, a request to deliver a target content to a target object is received. For a content delivery strategy of a plurality of content delivery strategies, at least one group of observed samples are determined based on target features of the target content and the target object, an observed sample comprising reference features of a reference content and a reference object, an indication of whether the content delivery strategy is applied to deliver the reference content and an object response. An effect metric is determined for the content delivery strategy based on the at least one group of observed samples. A target content delivery strategy for delivering the target content to the target object is selected from the plurality of content delivery strategies based on respective effect metrics for the plurality of content delivery strategies.

There are provided methods, devices, and computer program products for video generation. A first video is converted into a second video to change the appearance of the first video. The first video comprises a first object that performs an action and the second video comprises the same object that performs the same action. A video generating model is determined based on the first and second videos. Specifically, A first feature is determined for the first video and a second feature is determined for the second video by the video generating model. A first motion portion in the first feature is determined for the first video based on a temporal change in the first video; the second feature is updated by aligning a second motion portion in the second feature with the first motion portion; the diffusion model is updated based on the first feature and the updated second feature.

There are provided methods, devices, and computer program products for image generation, particularly to debiasing text-to-image diffusion models. In a method, a plurality of images is obtained by an image generating model based on a prompt. The plurality of images comprises a plurality of instances of an object, respectively and the object is specified by the prompt. A plurality of attributes of the plurality of instances of the object are determined respectively. The image generating model is updated based on the plurality of attributes and a predetermined distribution of a plurality of predetermined attributes related to the object. With the above method, the images generated by the updated image generating model may follow the predetermined distribution, and the updated image generating model may output debiased results.

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for synthetic data generation for service recommendation models. One of the methods includes generating a plurality of synthetic data items, each synthetic data item including user data, context data, service data of a service, and user action data characterizing one or more user actions with respect to the service; for each synthetic data item of the plurality of synthetic data items, processing the synthetic data item using an evaluator to generate a score indicating quality of the synthetic data item; selecting a subset of synthetic data items from the plurality of synthetic data items; and providing the subset of synthetic data items into a service recommendation model for recommending one or more services to one or more users based on the subset of synthetic data items.

Implementations for a space-optimized graph database system are provided. One implementation includes a computing system comprising: processing circuitry and memory storing instructions that causes the processing circuitry to: store a graph database comprising an initial tree graph storing a plurality of data entries, each data entry comprising a respective field identifier; receive a query to update the graph database, wherein the query comprises a request to add a new data entry; determine a splitting event to perform based on one or more predetermined criteria; generate a new tree graph corresponding to a field identifier of the new data entry by splitting off a subset of the plurality of data entries of the initial tree graph, wherein the subset comprises all data entries of the initial tree graph that correspond to the field identifier of the new data entry; and update the new tree graph in accordance with the query.

Implementations for I/O efficient synchronization on a graph database (102) are provided. One example includes a computing system (100) comprising: processing circuitry (104) and memory (106) storing instructions that, during execution, causes the processing circuitry to: implement the graph database (102) using at least: a read-write node (110) and a read-only node (112) operating in cache memory; and a shared storage (114) for data persistence, wherein the shared storage (114) stores a mapping table (202) of the graph database (102); receive an update query; update the read-write node (110) based on the update query; write, using the read-write node (110), a set of logs corresponding to the update query to a write-ahead log journal in the shared storage; and retrieve, using the read-only node (102), the set of logs from the write-ahead log journal.

Implementations for performing workload-aware space reclamation are provided. One example includes a computing system comprising: processing circuitry and memory storing instructions that, during execution, causes the processing circuitry to: track extent usage characteristics for each of a plurality of extents storing data for a graph database; and perform memory reclamation to reclaim storage areas corresponding to invalid data, wherein performing the memory reclamation comprises: generating a list of extents to be reclaimed based on the tracked extent usage characteristics; and reclaiming memory space on the plurality of extents in order based on the list of extents.

Described herein is an implementation of a classification of content collection using process supervision with a chain-of-thought reasoning capable machine-learned (ML) model. In an example aspect, a process supervision system includes processing circuitry and memory storing a content source of a ML model. The content source includes text-labeled content of a first type. The processing circuitry is configured to obtain a collection of multiple first-typed content items of the content source. The collection has text metadata associated therewith. The processing circuitry is further configured to produce supervised fine-tuning (SFT) queries regarding the collection based, at least in part, on one or more of the first-typed content items of the collection and/or the text metadata of the collection. The processing circuitry is still further configured to, based on the SFT queries, assemble a chain-of-thought reasoning request for submission to a chain-of-thought reasoning capable ML classifier.

Embodiments of the present disclosure provide a method and an apparatus for protecting data in an electronic device, an electronic device, and a program product. The method (200) includes: receiving (202) first data from a first application by a security service, where the security service blocks the first application from accessing a second application to protect data. The method further includes: receiving (204) second data from the second application by the security service. In addition, the method further includes: determining (206) third data for the electronic device by the security service based on the first data and the second data.