Abstract: A disaggregated-memory-based learned index write extension method and system are provided, wherein the system includes compute-pool servers and memory-pool servers. The compute-pool server holds a pointer to a write delta buffer of the reuse model, determines insertion location of key-value pairs based on a linear model in cached learned index model, and uses the held pointer to access a write delta buffer on the memory-pool server. The memory-pool server constructs the learned index model and the write delta buffer of reuse model to absorb newly inserted key-value pairs. A retraining thread in the memory-pool server asynchronously retrains submodels, scans and trains key-value pairs within the submodel, generates new bottom-layer submodels, updates the write delta buffer, and dynamically adjusts size of the write delta buffer based on data insertion characteristics within the coverage range of the submodel.

Abstract: A method and apparatus for accelerating KV-separated LSM-tree storage indexing based on smart SSD are disclosed. The method comprises: constructing, on a host side, a KV-separated LSM-tree storage module comprising a GC manager and a GC scheduler, wherein GC manager constructs and manages a ValidMap to record KV data validity at respective positions in a value file, and GC scheduler is configured to initiate, schedule, and write back results of GC compute units; constructing GC compute units on a smart SSD side to offload a garbage collection module from the host side, the GC compute units performing garbage collection via decoding, processing, and encoding without data dependencies; and determining a parameter configuration for the number of deployed GC compute units meeting throughput requirements.

Abstract: The present invention relates a method and a system for cross-chain consensus oriented to federated learning, comprising: conducting intra-cluster single-chain federated learning and collecting local update information; sending updates after consensus to a second federation so as to execute cross-cluster gradient exchange; receiving a verification result of cross-cluster gradient update consensus fed back from the second federation; and conducting local model update based on the verification result. After implementation of the update consensus, the present invention provides rewards and punishments based on the contributions of the cluster representatives, thereby encouraging the cluster representatives in the computing nodes to act honestly, so that the participants can actively help the model update.

Abstract: An acceleration method for heterogeneous graph neural networks based on meta-path graphs is provided, including: constructing a meta-path graph by arranging all meta-path instances in graph form, based on a given heterogeneous graph and specified types of meta-paths; partitioning the meta-path graph to obtain multiple meta-path subgraphs, which are further divided by workload; performing layer-based encoding on the meta-path instances according to the workload distribution; merging all meta-path instance slice encodings based on inter-layer relationships in the meta-path graph to obtain meta-path instance encodings; and performing intra-meta-path aggregation and inter-meta-path aggregation to compute the final features of the target vertices. The present method significantly reduces redundant computations in heterogeneous graph neural network, thereby improving model inference efficiency.

Abstract: A method and system for encryption and assured deletion of information is provided, the method at least includes: sorting fields of the information into at least two sensitivity levels by sensitivity; generating encryption keys and key shards thereof based on predetermined thresholds, and creating mapping between targets and the key shards, based on the encryption keys for the sensitivity levels, encrypting the information fields of the corresponding sensitivity levels and deleting the original information and encryption keys; and in response to reception of a recover request, recovering the encryption keys based on the key shards and performing decryption, so as to recover the original information. The present disclosure aims at the problem that information is difficult to be safely stored and assuredly deleted, and realizes multi-party security key deletion of encrypted personal information.

Abstract: A method for incremental metapath storage and dynamic maintenance is provided, which includes, reformatting metapath instances, from a designated heterogeneous graph and of a designated metapath type, into path graphs; executing graph updating tasks and performing dynamic maintenance on the updated path graphs, traversing the path graph to obtain the location of metapath updates and update the path graph; for metapaths with length greater than 2 and with symmetrical central portion, central merge operation is performed to simplify path graph and perform subsequent restoration operation; and directly perform restoration operation on path graphs that do not meet the merging conditions. The present disclosure utilizes characteristics of graph update to obtain locality of metapath updates, and combines internal relationship characteristics of metapath instances to greatly speed up metapath generation and achieve real-time inference of dynamic heterogeneous graph models.

Abstract: The present invention relates to an interference-based method for knowledge graph completion and system thereof, wherein the method at least comprises: when performing sampling on a knowledge graph, constructing a knowledge graph completion model; performing model training and performance evaluation on the knowledge graph completion model; and performing prediction on missing elements of incomplete triples in the knowledge graph; the knowledge graph completion model is constructed through: based on optical interference and superposition principles, constructing a score function from data of superposed luminous intensities, mirroring the triples in the knowledge graph to a process of superposition of the luminous intensities, and differentiating between positives and negatives obtained during the sampling of the knowledge graph.

Abstract: Systems and methods are provided for developing/updating training datasets for traffic light detection/perception models. V2I-based information may indicate a particular traffic light state/state of transition. This information can be compared to a traffic light perception prediction. When the prediction is inconsistent with the V2I-based information, data regarding the condition(s)/traffic light(s)/etc. can be saved and uploaded to a training database to update/refine the training dataset(s) maintained therein. In this way, an existing traffic light perception model can be updated/improved and/or a better traffic light perception model can be developed.

Abstract: Systems and methods of optimizing vehicle trace data for generating digital representations of road networks are provided. For example, a methodology of the presently disclosed technology may comprise: (1) segmenting vehicle trace data into a first set of tile groups; (2) applying an optimization algorithm to the first set of tile groups; (3) segmenting the vehicle trace data into a second set of tile groups, wherein geospatial arrangement of the second set of tile groups is shifted with respective to geospatial arrangement of the first set of tile groups; (4) applying the optimization algorithm to the second set of tile groups; and (5) generating a representation of an environment based on the application of the optimization algorithm to the first and second sets of tile groups.

Abstract: A method and system for deleting multi-copy personal data efficiently and securely is provided, wherein the personal data and its subject identifier are signed and uploaded to data domains and stored as personal data copies; the personal data copies along with its source and destination data are circulated among the data domains; the data domain receiving a deletion instruction transmits the deletion instruction to every relevant data domains based on the identifier of the personal data subject and the destination data and then performs deletion; and after completing the deletion, the data domain deposit its domain identifier and feedback data it receives into a log, and feed the log back to its superior data domain. And the system of the present disclosure includes a plurality of data domains that can perform the above operations, thereby realizing association-based storage, association-based deletion and verification of association-based deletion of multi-copy personal data.

Abstract: The present invention relates to a post-exascale graph computing method, and corresponding system, storage medium and electronic device. The invention solves the problems of low computing performance, poor scalability and high communication overhead in the large-scale distributed environment, and improves the performance of the supercomputer when supporting large-scale graph computing.

Abstract: The present invention relates to a graphic-blockchain-orientated sharding storage apparatus, at least comprising a first sharding module and a second sharding module, wherein the first sharding module shards nodes having different resource capacity levels based on ledger data organized using a DAG structure, and the second sharding module assigns transactions to the shards matching with execution difficulty levels of the transactions, so that each said transaction is processed and stored in the shard corresponding thereto. The present invention incorporates the sharding technology into a graphic blockchain to provide a graphic-blockchain-orientated sharding storage method, so as to reduce pressure in terms of data storage and transaction processing on nodes of the graphic blockchain system.

Abstract: Systems and methods are provided that can improve a SLAM system to account for the inaccuracy in the timestamps of location data and image data that operate independently or with dedicated control units. For example, the systems and methods may replace the timestamps with a counter value that is communicatively coupled with the vehicle to generate an incremental progression of the data generated by the sensors. Additionally, the process may model data points (e.g., landmarks) in accordance with an uncertainty (e.g., covariance) that is shaped as an ellipsoid that represents the uncertainty in the position of the vehicle at a point based on the location sensor. As the speed of the vehicle increases, in one approach, the more elongated the ellipsoid becomes. The uncertainty in the time of the location data may be modeled as a covariance so that the location data can be modeled through the SLAM system.

Abstract: Systems and methods are provided for verification of the mapping output while accounting for the large amount of data associated with a full geometry map. For example, the system may generate the geometry map of an environment where a vehicle is located and automatically identify a defective area of the geometry map. The defective area may, for example, be identified using a machine learning model to detect the defective area with respect to a threshold or confidence value. The system can receive a bounding box from at least one user device that identifies an adjustment to the defective area of the geometry map. Using the bounding box, the system can crop the defective area of the geometry map and initiate an action based on the adjustment to the defective area of the geometry map.

Abstract: Systems, methods, computer-readable media, techniques, and methodologies are disclosed for performing end-to-end, learning-based keypoint detection and association. A scene graph of a signalized intersection is constructed from an input image of the intersection. The scene graph includes detected keypoints and linkages identified between the keypoints. The scene graph can be used along with a vehicle's localization information to identify which keypoint that represents a traffic signal is associated with the vehicle's current travel lane. An appropriate vehicle action may then be determined based on a transition state of the traffic signal keypoint and trajectory information for the vehicle. A control signal indicative of this vehicle action may then be output to cause an autonomous vehicle, for example, to implement the appropriate vehicle action.

Abstract: A cache-designing method using cache lines to record cache-miss information is provided, wherein cache lines and cache-miss information are stored in a common storage space by means of shared storage. Tags of cache lines, as well as cache lines and cache-miss information, are stored separately in different static random-access memories, wherein multiple independent memories are used for tags, while a single memory is for cache lines and cache-miss information. A request-processing pipeline and a response-processing pipeline are constructed to be parallelable and used respectively for processing memory-access requests and for processing memory-response data.

Abstract: A method and system for kernel data isolation based on multiple kernel page tables is provided. Through creating multiple kernel page tables in a kernel system and binding specific applications to a corresponding page table for operation, each internal process of the application remains consistent with the corresponding page table in terms of kernel address space, and the kernel address spaces of different applications are isolated from each other due to existence of the different page tables. The system includes a page table management module and a private memory management module. The present disclosure protects private application data at the granularity of applications in the same kernel at very low costs to prevent data breach caused by read attacks on kernel address space or privilege escalation attacks, maintains privilege level division of kernel address spaces, and ensures transparency of private memories among applications and general IO capability of private memories.

Abstract: A GPU-sharing method and apparatus for serverless inference loads is provided, wherein the method involves intercepting and forwarding GPU API calls made by inference tasks to an API proxy process to manage and allocate GPU resources. With a CPU and multiple GPUs connected through a bus, the GPUs communicate with the CPU only through an API proxy for process management and resource allocation of the GPUs. The CPU intercepts all GPU APIs triggered by any function of a same inference application, forwards the intercepted GPU APIs to a same designated GPU runtime for execution, and directs the GPU APIs triggered by each function to a pre-designated stream pool for the same inference application, so that all the functions of the same inference application share the same GPU runtime. The present disclosure solves the problem related to bulkiness of GPU runtimes in serverless inference systems, thereby facilitating GPU resource usage.

Abstract: A sample-difference-based method and system for interpreting a deep-learning model for code classification is provided, wherein the method includes a step of off-line training an interpreter: constructing code transformation for every code sample in a training set to generate difference samples; generating difference samples respectively through feature deletion and code snippets extraction and then calculating feature importance scores accordingly; and inputting the original code samples, the difference samples and the feature importance scores into a neural network to get a trained interpreter; and a step of on-line interpreting the code samples: using the trained interpreter to extract important features from the snippets, then using an influence-function-based method to identify training samples that are most contributive to prediction, comparing the obtained important features and the most contributive training samples, and generating interpretation results for the object samples.

Abstract: The present invention relates to a blockchain-based system and method for preventing unauthorized deletion of surveillance video, the system at least comprising: a plurality of camera components, for generating surveillance video; at least one video-recording node, for connecting to a blockchain network and randomly generating a key pair, splitting the video file received from the camera components into a plurality of video-file blocks and randomly sending them to participation nodes across the blockchain network for storage, and recording hash values of the stored video files and node information; and a blockchain network device, for updating the hash values of the video files uploaded by the video-recording nodes based on a smart contract. In the present invention, uploading and storing time for camera-videos is recorded by the smart contract as tamper-resistant time reference for law enforcement, protecting video data from malicious deletion at video-recording nodes without the need of redundant space.