352 results about "Improved performance" patented technology

A first neural network (NN) model may generate labels for training a second NN model. The second NN model may represent instances of a NN model operating on multiple different devices (e.g., decentralized user and / or edge devices). The system may include using a “teacher” model to process data received by one or more of the devices to generate a labeled dataset. The system may use the labeled dataset and a “student” model to calculate gradient data for updating the student model. The student model may be the same or similar to NN model instances operating on the devices. The system may validate the updated student model to determine, for example, whether it exhibits improved performance when processing the newly received data and / or historical data. The system may distribute the validated update to the devices.

A golf club head and golf club having large head dimensions, and large face characteristics, as well as unique mass property attributes driven by relationships not often considered in conventional club head design, to achieve a resistance to squaring the club head during a golf swing that is comfortable to the novice golfer, while increasing stability during off-center impacts and obtaining preferred launch characteristics. This is achieved in part via establishing a club head configuration and associated weight distribution to yield a center of gravity location that results in a preferred magnitude of Delta1 and Delta2 values, CG angle, moments of inertia, and associated ratios, relationships, and club head mass property characteristics influenced by these variables to achieve improved performance.

A method of serving a generative transformer model includes determining a batch size to use in processing inference requests and allocating at least one prompt pipeline and at least on token pipeline to the generative transformer model to process the batch of inference requests. The number of prompt pipelines and the number of token pipelines, and the depths of the pipelines are determined based on the batch size, an average prompt length, a cache requirement per stage, and a memory footprint of model weights for the generative model per stage using a resource allocator component of the model serving system. Cache streaming is used to stream prompt cache from prompt pipelines to token pipelines to generate tokens. Cache streaming involves gather-copy operations which may be performed using compute kernels.

A torque control fastener system having a fastener nut, a helical wire insert, and a shaft, where the nut and helical wire insert have unique hardness and coefficient of thermal expansion relationships that produce improved performance.

The invention relates to a brake pad or a disc brake caliper, comprising a support plate receiving a friction material. According to the invention, the support plate is made from low-alloy carbon steel containing less than 0.3% by weight of carbon and vanadium in an amount between 0.2 and 0.3% by weight.

A system is disclosed, including a display, a processor and a memory. The memory stores instructions that, when executed by the processor, configure the system to perform operations. Active region data is generated that includes, for each of one or more active regions, active region location data and active region content. The active region data is transmitted to a display having a display area. For each active region, the active region content is displayed at an active region location of the display area based on the active region location data, the active region content being displayed at a higher spatiotemporal information density than content displayed in the display area outside of the active regions.

The present technology, roughly described, provides for mocking an application program interface (API) using a large language model (LLM). The present system generates a prompt with API signature information and API desired behavior information. The prompt can include instructions, library functions, examples, executed programs, and a current function invocation, as well as other content. The prompt can be generated, and submitted to an LLM to mock an API and generate a response. The response can be audited and the LLM can be fine tuned to provide improved performance in subsequent calls.

An electronic device includes a multilevel package substrate, a semiconductor die, and a package structure, the multilevel package substrate has first, second, and third levels including respective dielectric layers and conductive features, the first level including a first trace layer with an antenna and a first via layer with a portion of a ground wall laterally spaced outward from and surrounding the antenna, and the second level including a second trace layer having a ground plane connected to the ground wall, the semiconductor die attached to the first level of the multilevel package substrate, and the package structure including a molding compound enclosing the semiconductor die and extending on a side of the antenna, where the package structure mold compound maters and thickness can be tuned for improved performance.

A radio frequency package for an electronic device with a reduced size and improved performance is presented herein. The radio frequency package includes a front-end package, transceiver dies and at least one antenna array providing a wireless communication functionality for the radio frequency package. The front-end package includes a radio frequency front-end die and at least one power control die connected to an insulation film substrate via first connectors. Each of the transceiver dies is connected to the at least one antenna array using second connectors.

A semiconductor structure having improved performance is provided that includes a local enlarged via-to-backside power rail (VBPR) contact structure which connects a source / drain region of one field effect transistor (FET) to a backside power rail.

A compact vane-axial fan with includes modular components and is associated with improved efficiency and can operate with a variety of differently configured impellers and vanes to allow for customization to a particular application. The fan can include a base set of components, including a fan housing with motor and stator vane. The base set of components can be modularly assembled with an impeller and a vane insert, which is positioned inline between the impeller and the stator. The configuration(s) of the impeller, the vane insert, and / or the stator contribute to a compact design. In addition, configurations of the impeller and the vane insert can be tailored to one another to contribute to improved performance. Furthermore, a variety of different combinations of impellers and vane inserts can be selectively assembled with the base set of components to tailor fan operations for a use case and still achieve high efficiency.

A vehicle aerodynamic system with dual operating modes to modify downforce across varying speed ranges and driving conditions. The system comprises multiple fans positioned in airflow pathways and deployable skirts that interact with the ground surface. In a first mode, a complete set of skirts creates a fully sealed bounded region under the vehicle with central fans generating maximum downforce at low speeds on smooth surfaces. In a second mode, a subset of skirts (primarily side skirts) modifies the bounded region while all fans operate to provide downforce in dynamic driving conditions with uneven surfaces. The system includes a control mechanism to selectively deploy skirts and operate fans based on driving conditions, optimizing between maximum downforce at low speeds and improved performance over varying terrain at higher speeds.

A vehicle, and in particular an offroad vehicle, with improved capabilities. The vehicle provides improved performance (e.g., ground clearance, deployment options, or the like), allows for customization for different types of vehicles (e.g., frame allows for the connection of different types of components for different applications, improved assembly, and / or maintenance), and / or allows for modular interchangeability of vehicle beds in the field for changing situations (e.g., different beds may be interchanged for different operational purposes).

The invention discloses a performance self-improving large model recommendation system and method, and the system comprises an initialization module which is used for pre-training a large language model through supervision and fine tuning, and generating an initial recommendation model; the self-optimization module comprises three iteratively executed sub-modules; the sample selection sub-module is used for screening historical data samples of which the information amount is higher than a threshold value on the basis of comparison between the model prediction probability and the preset threshold value; the response fusion sub-module is used for generating K candidate responses for a selected sample and generating a preference data set based on model evaluation; and the DPO optimization sub-module is used for updating model parameters by utilizing the improved DPO loss function and generating an optimized recommendation model. According to the method, the dependence on static preference data in the prior art is broken, the recommendation quality and robustness are improved, and adaptive optimization is realized.

Systems and methods for unified federated learning for time series forecasting applications are described herein. In one example, an edge server includes a processor and a memory in communication with the processor. The memory includes instructions that, when executed by the processor, cause the processor to train a model deployed on the edge server using information from one or more external devices received by the edge server and in response to a determination that the training of the model improved performance of the model, deploy the model. In addition, the instructions also cause the processor to, in response to a request from an aggregation server, transmit one or more model parameters of the model to the aggregation server and, in response to receiving an updated model from the aggregation server, deploy the updated model.

A torque control fastener system having a fastener nut, a helical wire insert, and a shaft, where the nut and helical wire insert have unique hardness and coefficient of thermal expansion relationships that produce improved performance.

To provide a game machine with improved performance control. [Solution] A gaming machine comprising a plurality of speakers of different types, a plurality of audio circuits electrically connected to each of the plurality of speakers and capable of outputting audio signals, and a first board on which the plurality of audio circuits are arranged, wherein each of the plurality of audio circuits comprises a first component, a plurality of second components, and a third component, and in each of the plurality of audio circuits on the first board, the positional relationship between the first component, the plurality of second components, and the third component is approximately the same, and in each of the plurality of audio circuits on the first board, the spacing between the plurality of second components is approximately the same, and no components are mounted in the area consisting of the spacing.

The invention provides a pathological image analysis method based on cross-scale spatial constraint fusion and local perception. The method comprises the steps of 1, data preprocessing and feature extraction; step 2, executing a two-way fusion strategy based on spatial constraint to obtain fused enhanced features; step 3, capturing high-order semantic information between scales by using a bilinear fusion mode based on a Hadamard product; 4, performing residual connection on the enhanced features and the fused features to serve as input of a 10-time branch and a 20-time branch; 5, remodeling the features after residual connection into a two-dimensional shape, and inputting three convolutional layers which are arranged in a layered manner and expanded layer by layer; and 6, further performing gating adaptive fusion on the global semantic features obtained after the 10-time branches and the 20-time branches are subjected to the step 5. On the basis of easy deployment, feature representation with high expression ability is obtained, information between branches of different scales is fully utilized, and better performance is obtained in downstream tasks.

To provide a game machine with improved performance control. [Solution] A gaming machine comprising a plurality of speakers of different types, a plurality of audio circuits electrically connected to each of the plurality of speakers and capable of outputting audio signals, and a first board on which the plurality of audio circuits are arranged, wherein each of the plurality of audio circuits comprises a first component, a plurality of second components, and a third component, and in each of the plurality of audio circuits on the first board, the positional relationship between the first component, the plurality of second components, and the third component is approximately the same, and in each of the plurality of audio circuits on the first board, the first component is arranged in an intermediate position between the plurality of second components.

An improved silicon carbide (SiC) planar MOSFET having at least one buried P-shield (BPS) region including one center portion and two side portions for the gate oxide electric field and saturation current reductions is disclosed. Two saturation current pitching (SCP) structures formed in two Junction Field Effect Transistor (JFET) regions sandwiched between the side portions and the center portion of the BPS region limit a saturation current of the device in a forward conduction stage for the short-circuit capability improvement. Moreover, a Junction barrier Schottky diode (JBSD) is integrated with the SiC MOSFET in a location between the two adjacent split gate electrodes for the reverse conduction switching loss reduction.

Artificial intelligence driven enhancement of motion blurred sequencing images enables enhanced sequencing that determines a sequence of bases in genetic material with any one or more of: improved performance, improved accuracy, and / or reduced cost. A training set of images taken after unreduced and reduced movement settling times during sequencing is used to train a neural network to enable the neural network to recover enhanced images, as if taken after the unreduced movement settling time, from unenhanced images taken after the reduced movement settling time.

Embodiments of the present disclosure disclose apparatuses, methods, and computer programs for acoustic echo cancellation (AEC) based on one or more diagonal regularized correlation matrices. At least some example embodiments may allow for an intelligent and efficient solution for multi-channel AEC and / or stereo AEC, providing improved performance and simpler implementation.

Described herein are techniques for analyzing historic time-series data to determine right-size commissioned resources from a software provider such as a hyperscaler. The historic time-series data may be analyzed to determine whether a commissioned resource should be upsized or downsized based on historical usage of the resource. Advantages to right-sizing a commissioned resource include improved performance and reduced spending.

The invention relates to a linear lighting unit (1) having a longitudinal extension (L1) and arranged to provide unit light. The linear lighting unit (1) comprises a first linear elongated light source (2) having a longitudinal extension (L2) substantially parallel to a longitudinal extension (L1) of the linear lighting unit (1). The first linear elongate light source (2) comprises a plurality of first LEDs (4) arranged along a longitudinal extension (L1) of the first linear elongate light source (2). The linear lighting unit further comprises a second linear elongate light source (3) having a longitudinal extension (L3) substantially parallel to the longitudinal extension (L1) of the linear lighting unit (1). The second linear elongate light source (3) comprises a plurality of second LEDs (5) arranged along a longitudinal extension (L3) of the second linear elongate light source (3). The plurality of first LEDs (4) are arranged to emit a first light beam (4 ') in a first direction (A). The second linear elongate light source (3) is preferably arranged at least partially within the first light beam (4 ') and is arranged to emit a second light beam (5') in a second direction (B).

Distance calculations for autonomous vehicle (AV) or advanced driver assistance system (ADAS) navigation may include calculating a distance between two points within an image captured by a vehicle camera. To calculate distance in ADAS and AV systems, the vectors and other values may be represented as floating-point (FP) values. The use of FP values may provide improved performance of these calculations in ADAS, AV systems, and other systems that seek to reduce or minimize computation speed. When the input values are represented as FP values, the distance may be calculated as a square root of a sum of squares of the two FP values. Improved systems and methods are provided for floating-point calculation related to a square root, such as for determining distance calculations.

A semiconductor structure and a method for forming the same, the method comprising: providing a substrate; forming a gate structure on the substrate; forming a first recess in the substrate on both sides of the gate structure, a bottom surface of the first recess being a concave arc surface; performing a modification treatment on a sidewall surface of the first recess; forming a second recess and a modification layer on a sidewall surface and a bottom surface of the second recess, a thickness of the modification layer at a corner of the sidewall surface and the bottom surface of the second recess being greater than a thickness of the modification layer on a sidewall of the second recess; and removing the modification layer, a bottom surface of the second recess being parallel to a surface of the substrate. The semiconductor structure formed by the method has improved performance.

This disclosure relates to a distribution matching circuit for data transmission. The circuit comprises a first stage distribution matcher (DM) bank and a second stage DM bank, connected via a crossbar interconnect. The first stage DM bank includes a plurality of first stage distribution matchers, which perform an initial stage of distribution matching, generating output bits that are then transmitted to the second stage DM bank via the crossbar interconnect. The second stage DM bank receives these output bits and performs a second stage of distribution matching. The circuit allows for adjustable input lengths in both stages, achieving flexible rates. The second stage DM outputs sequences of different costs. This disclosure is particularly suited for optical communication systems requiring fine-grained rate flexibility and improved performance.

A tire is provided that can provide improved performance on snow without compromising wear resistance. A tread portion (1) is provided with at least three curved main grooves (10) extending in a zigzag along a circumferential direction of the tire. A land portion (20) between the curved main grooves (10) is defined by a plurality of lug grooves (30) in a plurality of blocks (20B). A road contact surface of each of the plurality of blocks (20B) is provided with a pair of sipes (40) and a pair of recess grooves (50). The pair of sipes (40) extend in the same direction within a range of an angular difference within 10°. Each of the pair of sipes (40) has one end open to the curved main grooves (10), and the other end blindly terminates in a block (20B) of the blocks (20B).The pair of recessed grooves (50) extend in the same direction within a range of angular difference within 10°. Each of the pair of recessed grooves (50) has one end open toward the curved main grooves (10), and the other end blindly terminates in the block (20B).

To provide a game machine with enhanced performance control.SOLUTION: A game machine includes: multiple speakers; multiple audio circuits connected to the multiple speakers; and a first substrate on which the multiple audio circuits are arranged. On at least a first audio circuit and a second audio circuit of the multiple audio circuits, a first component, multiple second components, and a third component are in the same or substantially the same positional relation. A certain connector electrically connectable to a certain speaker is arranged near a certain audio circuit. The first component, the second component, and the third component are, respectively, an audio amplifier, a coil, and a capacitor. The first audio circuit and the second audio circuit include filter circuits consisting of the coil and the capacitor. The first substrate is provided with thermal vias corresponding to the audio amplifier, and the thermal vias are arranged in the same or substantially the same manner in the first audio circuit and the second audio circuit.SELECTED DRAWING: Figure 27

The present application relates to a method for generating communication data processing code, a computer device and a medium, including: obtaining the interface information of an RPC function; recursively extracting the first named structures associated with the input and output of the RPC function and removing duplicates based on the interface information; generating a Request structure based on the input of the RPC function, and generating a Response structure based on the output of the RPC function; obtaining a package associated with the first named structure, the Request structure and the Response structure; if the package lacks a cache file or the cache file is updated, determining the corresponding target structure set based on the several second named structures included in each package; sorting the second named structures in the target structure set respectively to obtain a sorted list of target structures corresponding to the package; and generating the target code corresponding to the package in sequence based on the sorted list of target structures. The code for serialization and deserialization generated based on the solution provided by the present application has greatly improved performance compared to traditional reflection solutions.