Abstract: A training method and an enhancement method for a speech enhancement model, an apparatus, an electronic device, a storage medium, and a program product are described. A training method for the model may include: invoking the model based on a noisy speech feature of a speech signal, to obtain first predicted mask values in an auditory domain. A first amplitude and first phase of the noisy speech signal, and a second amplitude and second phase of a clean speech signal may then be obtained. A phase difference at each frequency point may be determined based on the phases. The second amplitude corresponding to each frequency point may be corrected based on the phase difference at each frequency point, to obtain a corrected second amplitude corresponding to each frequency point. A loss value is determined, and parameters of the speech enhancement model may be updated based on the loss value.

Abstract: A spray pipe assembly includes a pipe body including a mounting channel configured to mount a water pipe. The spray pipe assembly also includes a flushing device, removably mounted at a front end of the pipe body. The flushing device includes a device main body including a mounting cavity. The flushing device also includes a flushing nozzle configured to clean a human body. At least a portion of the flushing nozzle is disposed in the mounting cavity. The flushing nozzle includes at least one spray hole facing upwards. The flushing device also includes a flushing spray head configured to clean a toilet cavity and connected to the device main body. The flushing spray head includes a spray head facing downwards.

Abstract: A die-casting aluminum alloy without heat-treatment and a preparation method and application thereof. Based on a total weight of the die-casting aluminum alloy, the die-casting aluminum alloy includes: 6.0 to 8.0 wt % of Si; 0.3 to 1.2 wt % of Mg; 0.4 to 0.8 wt % of Cu; 0.1 to 0.3 wt % of Fe; 0.6 to 0.8 wt % of Mn; 0.05 to 0.20 wt % of Ti; 0.03 to 0.07 wt % of Sr; 0.03 to 0.07 wt % of Ce; 0.01 to 0.04 wt % of La; 0.01 to 0.1 wt % of Zr; less than or equal to 0.01 wt % of other impurity elements; and a balance of Al.

Abstract: The disclosure provides an optical anti-counterfeiting element, an optical anti-counterfeiting product and a manufacturing method. The optical anti-counterfeiting element includes: a substrate; a micro-structure forming layer, wherein the micro-structure forming layer is arranged on a side surface of the substrate, the micro-structure forming layer has micro-structures, and a longitudinal section of each of the micro-structures has a preset shape; and a coating layer, wherein the coating layer is arranged on the micro-structure forming layer, the micro-structures are coated with the coating layer to form a coated micro-structure, and a height of each of the coated micro-structures continuously changes; wherein when shapes of the micro-structures have an abrupt-change, the coating layer has a demetallized region, and when a height of each of the micro-structures changes continuously, the coating layer has no demetallized region.

Abstract: Apparatuses, systems, and techniques are presented to perform segmentation on images. In at least one embodiment, one or more neural networks are used to segment an image based, at least in part, on one or more visual modifications of the image.

Abstract: Apparatuses, systems, and techniques to train one or more neural networks based, at least in part on, medical imaging data and clinical metadata or inference using one or more neural networks trained as such. In at least one embodiment, one or more circuits to train one or more neural network to predict a treatment for a patient suspected to have or confirmed to have COVID-19 based, at least in part on, medical imaging data and clinical metadata.

Abstract: Apparatuses, systems, and techniques to train neural networks to perform classification. In at least one embodiment, one or more neural networks are trained to perform classification based on, for example, using one or more compressed representations of one or more class labels, where the one or more compressed representations have fewer bits than a representation of the one or more class labels.

Abstract: Disclosed is a Bayesian classification recognition system based on an industrial PaaS platform, comprising: an IaaS infrastructure service layer, a G-PaaS graph neural network processing layer, an O-PaaS docking service layer and an SaaS system application layer. The G-PaaS graph neural network processing layer is configured for point cloud feature generation, point cloud feature learning, point cloud structure estimation and point cloud model classification; and the recognition accuracy of a workpiece point cloud model is improved through the Bayesian classification recognition system based on the industrial PaaS platform.

Abstract: NKG2D-CD3 and NKG2D-CD16 cell adapter molecules are used in the removal of aging cells and the treatment of aging-cell-accumulation-related diseases. The cell adapter molecules has comprise an NKG2D extracellular domain, a connection segment, and a CD3 or CD16 binding domain. The death of aging cells having high expression of NKG2D ligand is efficiently induced by means of bridging aging cells and immune cells, and thus the cell adapter molecules can be used for removing aging cells expressing the NKG2D ligand and treating aging-related diseases.

Abstract: An NK cell adapter molecule targeting an NKG2D ligand and an NKp46 and the use thereof are provided. The cell adapter molecule at least comprises: a first binding domain comprising an NKG2D extracellular domain that can specifically bind to the NKG2D ligand; and a second binding domain comprising an NKp46 antigen-binding fragment. The cell adapter molecule can promote the killing effect of natural killer cells on target cells with high levels of expression of NKG2D ligand, and can thus be used as a therapeutic agent for diseases related to NKG2D ligand expression.

Abstract: Apparatuses, systems, and techniques to infer synthesized images including digital representations of objects blended realistically with appropriate background images. In at least one embodiment, background image data and a description of semantic features for a type of object are fused together to infer such a synthesized image using one or more neural networks.

Abstract: A battery pack includes: a sensor that generates sensor data indicating an operating parameter of the battery pack; and an electronic controller including a machine learning program. The electronic controller is configured to: receive the sensor data; process the sensor data using the machine learning program, where the machine learning program includes a trained neural network model; and use the machine learning program to generate an output based on the sensor data, where the output indicates at least one of a state of charge (SOC), a state of temperature (SOT), a state of health (SOH), and a state of power (SOP) of a cell. The effective utilization of a battery is improved.

Abstract: A method for preparing lithium bis(fluorosulfonyl)imide, includes reacting bis(fluorosulfonyl)imide with lithium carbonate in a mixed organic solvent to obtain lithium bis(fluorosulfonyl)imide after post-treatment. In the synthetic route of the lithium bis(fluorosulfonyl)imide, the reaction process stays at the stage of lithium bicarbonate, so there is little water in the reaction system, and it is very easy to recover the generated lithium bicarbonate. The process is simple and can be used to prepare lithium bis(fluorosulfonyl)imide of high quality at a high yield.

Abstract: Apparatuses, systems, and techniques to indicate an extent, to which text corresponds to one or more images. In at least one embodiment, an extent to which text corresponds to one or more images is indicated using one or more neural networks and used to train the one or more neural networks.

Abstract: Automatic volumetric quantification can be performed for various parameters of an object by providing volumetric data, such as three-dimensional image data, to at least one neural network. A network can extract features from the data that can be used to infer a point cloud representative of the surface of the object. One or more loss functions can be used to adjust the relevant network parameters. The network can also attempt to infer a segmentation mask for the object, indicating which data values correspond to the object of interest. Since the network performs the segmentation and point cloud generation in parallel, updates to the network parameters can impact the segmentation process, effectively constraining the segmentation based on the inferred shape of the object. Ensuring that the segmentation mask corresponds closely to the surface of the object can cause the segmentation process to be more accurate than conventional segmentation processes alone.

Abstract: Apparatuses, systems, and techniques to select a neural network architecture from a plurality of neural networks in a federated learning (FL) setting. In at least one embodiment, a neural network is trained by combining training results from different FL computing systems, where each of the different FL computing systems, for example, trains different portions of the neural network.

Abstract: Apparatuses, systems, and techniques are presented to predict annotations for objects in images. In at least one embodiment, one or more annotations corresponding to one or more objects within one or more images are generated based, at least in part, on one or more neural networks iteratively trained using the one or more images.

Abstract: In at least one embodiment, an object detection system uses a neural network to identify and/or locate a set of organs in a medical image. In at least one embodiment, when training to identify and/or locate a particular organ, a subset of incompletely-labeled training images is used that excludes training images for which labels associated with particular organ are unavailable.

Abstract: An adapter includes: a power input terminal for connecting an AC power supply; a power circuit that performs energy conversion on the inputted AC power to obtain electrical energy capable of powering a power tool; and a tool interface connected to a power interface of the power tool to output electrical energy. The adapter further includes a data transmission module connected to at least the tool interface and capable of outputting first communication data and/or first internal resistance data to the tool interface so that the power tool is capable of identifying the adapter according to the first communication data and/or the first internal resistance data. The first communication data includes at least adapter identification data, and the adapter identification data enables the power tool to identify that a connected device is the adapter between the adapter and a battery pack.

Abstract: Apparatuses, systems, and techniques to detect object in images including digital representations of those objects. In at least one embodiment, one or more objects are detected in an image based, at least in part, on points corresponding to a surface of one or more objects.