Abstract: Methods, apparatus, and processor-readable storage media for authenticating usage data for processing by machine learning models are provided herein. An example method includes receiving, by a machine learning application installed in a user space of an operating system of a user device, a message from a software component, wherein the software component is: configured to collect usage data associated with the user device; signed with using private key corresponding to a digital certificate by an application installed on the user device; and deployed in a kernel space of the operating system, and wherein the message comprises usage data signed using the private key; authenticating, by the machine learning application, the usage data based on a public key corresponding to the digital certificate; and processing, by the machine learning application in response to a result of the authenticating, at least a portion of the authenticated usage data.

Abstract: Embodiments of the present disclosure relate to a method, a device, and a computer program product for executing computer programs. The method includes implementing multiple executions of a deep neural network that includes a hooking portion outside a trusted execution environment (TEE), wherein an operator in the hooking portion, when executed outside the TEE, invokes a corresponding execution operator in the TEE. During the period, the method includes determining, on the basis of the invocation of the corresponding execution operator in the TEE by the operator in the hooking portion, a computation graph corresponding to the hooking portion. The method further includes executing, on the basis of the computation graph, the execution operator corresponding to the hooking portion in the TEE during the execution of the deep neural network after the multiple executions.

Abstract: Techniques for integrating a trusted execution platform with a function-based service framework are disclosed. For example, a method comprises reading configuration information identifying at least one data providing party and at least one function providing party, and generating, based at least in part on the configuration information, an enclave comprising a circuit configured to execute a function. The method further comprises receiving in the enclave and via at least a first secure communication channel, the function from the at least one function providing party, and receiving in the enclave and via at least a second secure communication channel, data from the at least one data providing party. The function and the data are sent to the circuit, wherein the circuit executes the function to compute at least one output based at least in part on the data.

Abstract: Embodiments of the present disclosure relate to a method, a device, and a computer program product for video compression. The method includes: segmenting, in response to one or more features of an object in a video having a periodic change, the video into a plurality of segments based on a cycle of the periodic change; and identifying focal regions in frames of the video that are associated with the object. The method further includes: compressing the video based on the plurality of segments and the focal regions. This solution provides a content-aware lightweight video compression solution that supports content-based video deduplication at multiple scales and breaks the spatio-temporal continuity constraint of video frames during compression, thus enabling more effective video compression.

Abstract: Embodiments described herein provide for systems and methods for implementing a neural network architecture for spoof detection in audio signals. The neural network architecture contains a layers defining embedding extractors that extract embeddings from input audio signals. Spoofprint embeddings are generated for particular system enrollees to detect attempts to spoof the enrollee's voice. Optionally, voiceprint embeddings are generated for the system enrollees to recognize the enrollee's voice. The voiceprints are extracted using features related to the enrollee's voice. The spoofprints are extracted using features related to features of how the enrollee speaks and other artifacts. The spoofprints facilitate detection of efforts to fool voice biometrics using synthesized speech (e.g., deepfakes) that spoof and emulate the enrollee's voice.

Abstract: Embodiments of the present disclosure relate to a method, a device, and a computer program product for video compression. The method includes: segmenting, in response to one or more features of an object in a video having a periodic change, the video into a plurality of segments based on a cycle of the periodic change; and identifying focal regions in frames of the video that are associated with the object. The method further includes: compressing the video based on the plurality of segments and the focal regions. This solution provides a content-aware lightweight video compression solution that supports content-based video deduplication at multiple scales and breaks the spatio-temporal continuity constraint of video frames during compression, thus enabling more effective video compression.

Abstract: A network congestion handling method is provided. The method is applied to an HPC cluster including a plurality of hosts, and the plurality of hosts are connected via a network device. A transmit end obtains a first acknowledgment message including a network congestion scenario identifier. The network congestion scenario identifier is an identifier indicating a scenario in which network congestion occurs in a process in which the transmit end transmits data to a receive end, and the transmit end and the receive end are any two hosts in the plurality of hosts. Then, the transmit end adjusts a data sending rate of the transmit end based on the network congestion scenario identifier.

Abstract: The present invention provides a zymolytic rice protein powder with low granular sensation, and a method for preparing the zymolytic rice protein powder comprising the following steps: (1) pretreatment of rice residue after sugar production; (2) high temperature extrusion pretreatment; (3) refining treatment; (4) wet crushing; (5) controlling zymolysis; (6) terminating zymolysis; (7) microjet treatment; (8) spray drying; (9) subsequent crushing. The protein content of the zymolytic rice protein powder in the present invention is more than 80%; the fat content thereof is less than 3%; the water content thereof is less than 5%; and the nitrogen solubility index thereof is more than 30%. The molecular weight distribution of the soluble protein is as follows: the proportion thereof with a molecular weight of less than 2000 Da is more than 70%.

Abstract: Embodiments of the present disclosure relate to a task processing method, a task processing device, and a computer program product. The method includes: receiving a task to be processed by a first neural network at a first computing node; determining that the first neural network failed to process the task; and sending the task to a second computing node for processing by a second neural network at the second computing node, wherein compared with the first neural network, the second neural network has at least one of the following: higher processing consumption and higher processing performance. With the technical solution of the present disclosure, a task can be processed quickly and accurately using a neural network with low processing consumption.

Abstract: Described herein are systems and methods for improved audio analysis using a computer-executed neural network having one or more in-network data augmentation layers. The systems described herein help ease or avoid unwanted strain on computing resources by employing the data augmentation techniques within the layers of the neural network. The in-network data augmentation layers will produce various types of simulated audio data when the computer applies the neural network on an inputted audio signal during a training phase, enrollment phase, and/or testing phase. Subsequent layers of the neural network (e.g., convolutional layer, pooling layer, data augmentation layer) ingest the simulated audio data and the inputted audio signal and perform various operations.

Abstract: Embodiments of the present disclosure relate to a method, an electronic device, and a computer program product for processing a video. The method includes acquiring a video, where the video includes at least a current frame and a previous frame that are adjacent to each other. The method further includes determining, based on a first pixel value of a pixel in the current frame and a second pixel value of a corresponding pixel in the previous frame, whether the current frame has changed relative to the previous frame. The method further includes determining availability of the current frame for a computer vision task if it is determined that the current frame has changed relative to the previous frame. With the method, video data that needs to be processed is reduced, the task load of a computing device is lowered, system power consumption is improved, and data processing efficiency is improved.

Abstract: Embodiments of the present disclosure include a video compression method, an electronic device, and a computer program product. In a video compression method provided by embodiments of the present disclosure, a video is segmented into multiple segments based on a first feature of an object extracted from the video, where the object has a first change in each of the multiple segments, and the first feature identifies a start state and an end state of the first change; the multiple segments are grouped based on a similarity of the first change of the object among the multiple segments; and the video is compressed based on the groups of the multiple segments. In this way, for a video with periodically repetitive content, the video is compressed by retaining a part of representative video segments, which can save storage space and network transmission resources and improve computation efficiency, thereby reducing computation cost.

Abstract: Disclosed are systems and methods including computing-processes executing machine-learning architectures for voice biometrics, in which the machine-learning architecture implements one or more language compensation functions. Embodiments include an embedding extraction engine (sometimes referred to as an “embedding extractor”) that extracts speaker embeddings and determines a speaker similarity score for determine or verifying the likelihood that speakers in different audio signals are the same speaker. The machine-learning architecture further includes a multi-class language classifier that determines a language likelihood score that indicates the likelihood that a particular audio signal includes a spoken language. The features and functions of the machine-learning architecture described herein may implement the various language compensation techniques to provide more accurate speaker recognition results, regardless of the language spoken by the speaker.

Abstract: A data processing method comprising: After receiving an ith Peripheral Component Interconnect Express (PCIe) packet, a network interface card stores a jth instruction segment in a jth storage unit that is in a first storage area. When all n instruction segments of a first send queue entry (SQE) are stored in the first storage area, the network interface card obtains the first SQE, an identifier of a queue pair (QP) to which the first SQE belongs, and a location identifier of the first SQE in the QP according to the instructions in n storage units in the first storage area; the network interface card performs data processing based on the identifier of the QP to which the first SQE belongs and the location identifier of the first SQE in the QP.

Abstract: A nanocomposite includes an oxygen vacancy-containing BiOX particle and a coating, where the coating is a biocompatible material. Under near-infrared (NIR) irradiation, the nanocomposite has a photothermal conversion efficiency of greater than or equal to 10%. Under NIR irradiation, the nanocomposite degrades 1,3-diphenylisobenzofuran (DPBF) at a rate of higher than or equal to 0.1 mmol/h. BiOX may be BiOF, BiOCl, BiOBr, BiOI, or BiOAt. A preparation method and a use of the nanocomposite are further provided. The nanocomposite is a bismuth oxyhalide nanomaterial with different numbers of oxygen vacancies and can be used for the photothermal therapy (PTT) of a tumor and for the integrated tumor diagnosis and treatment. The nanocomposite leads to an excellent therapeutic effect under the guidance of multi-modality imaging, and has excellent computed tomography (CT) imaging and photoacoustic imaging (PAI) performance.

Abstract: Embodiments include a computer executing voice biometric machine-learning for speaker recognition. The machine-learning architecture includes embedding extractors that extract embeddings for enrollment or for verifying inbound speakers, and embedding convertors that convert enrollment voiceprints from a first type of embedding to a second type of embedding. The embedding convertor maps the feature vector space of the first type of embedding to the feature vector space of the second type of embedding. The embedding convertor takes as input enrollment embeddings of the first type of embedding and generates as output converted enrolled embeddings that are aggregated into a converted enrolled voiceprint of the second type of embedding.

Abstract: A method for storage management comprises: obtaining, at a first storage service deployed at a first node, a first request of a first computing service deployed at the first node for first target data, the first storage service having access to a remote storage device and providing the first computing service with a same access interface as the remote storage device, the remote storage device storing a dataset reusable in a task to be performed at least partially by the first computing service, the dataset comprising the first target data; obtaining, based on the first request, the first target data from the remote storage device or a local storage space for the first storage service; and providing the first target data to the first computing service.

Abstract: Techniques for generating a network topology are provided. For example, a method comprises: sending, at an edge node of a network, first exploration data to a first node of the network. The method also comprises receiving from the first node first feedback information for the first exploration data. The first feedback information includes identity information of the first node and first associated node list information, and the first associated node list information includes identity information of nodes associated with the first node in the network. In addition, the method comprises: determining latency between the first node and the second node based on the first latency information and the second latency information. Then, the method further comprises updating the network topology based on the latency between the first node and the second node. Through the method, the present disclosure can provide accurate data support for path optimization.

Abstract: The embodiments execute machine-learning architectures for biometric-based identity recognition (e.g., speaker recognition, facial recognition) and deepfake detection (e.g., speaker deepfake detection, facial deepfake detection). The machine-learning architecture includes layers defining multiple scoring components, including sub-architectures for speaker deepfake detection, speaker recognition, facial deepfake detection, facial recognition, and lip-sync estimation engine. The machine-learning architecture extracts and analyzes various types of low-level features from both audio data and visual data, combines the various scores, and uses the scores to determine the likelihood that the audiovisual data contains deepfake content and the likelihood that a claimed identity of a person in the video matches to the identity of an expected or enrolled person.

Abstract: The present invention provides a method for producing a rice protein peptide including (1) preparing a slurry of rice residue protein, followed by sterilizing the rice residue protein; (2) first crushing; (3) undergoing a first proteolysis; (4) second crushing; (5) undergoing a second proteolysis; (6) conducting solid-liquid separation; (7) concentrating; (8) membrane filtering; (9) providing an anti-microbial treatment; (10) performing spray drying. The present method performing super-fine grinding on the proteolytic rice residue protein. During the process of second proteolysis, the protease fully contacts with the substrate, thereby promoting the proteolytic efficiency and taste.