摘要:
The interconnection chip of the present invention is a custom chip which is designed to serve as an efficient link between system functional modules, such as arithmetic units, register files and input/output ports. The chip includes a crossbar interconnection, a FIFO or programmable delay for each of its inputs and a pipeline register file for each of its outputs. By using pre-stored control patterns, the chip can configure its crossbar and delays while performing other operations. Therefore, the usual functions of busses and register files can be realized with this single chip. Various embodiments and applications for the chip are disclosed.
摘要:
Accelerator systems and methods are disclosed that utilize FPGA technology to achieve better parallelism and flexibility. The accelerator system may be used to implement a relevance-ranking algorithm, such as RankBoost, for a training process. The algorithm and related data structures may be organized to enable streaming data access and, thus, increase the training speed. The data may be compressed to enable the system and method to be operable with larger data sets. At least a portion of the approximated RankBoost algorithm may be implemented as a single instruction multiple data streams (SIMD) architecture with multiple processing engines (PEs) in the FPGA. Thus, large data sets can be loaded on memories associated with an FPGA to increase the speed of the relevance ranking algorithm.
摘要:
Some implementations provide techniques and arrangements to receive image information. A plurality of reference fields of a user-manipulated device may be identified. Each reference field of the plurality of reference fields may include reference elements. The plurality of reference fields may be identified based on colors of the reference elements, shapes of the reference elements, and/or a pattern of the reference elements. Some implementations may generate position information based on the plurality of reference fields. The position information may identify a position of the user-manipulated device relative to the video camera. Some implementations may provide the position information to an application.
摘要:
A method using a RankBoost-based algorithm to automatically select features for further ranking model training is provided. The method reiteratively applies a set of ranking candidates to a training data set comprising a plurality of ranking objects having a known pairwise ranking order. Each round of iteration applies a weight distribution of ranking object pairs, yields a ranking result by each ranking candidate, identifies a favored ranking candidate for the round based on the ranking results, and updates the weight distribution to be used in next iteration round by increasing weights of ranking object pairs that are poorly ranked by the favored ranking candidate. The method then infers a target feature set from the favored ranking candidates identified in the iterations.
摘要:
Architecture for decoding (demosaicing) a source image and performing reconstruction directly from the Bayer pattern to reduce memory size and improve communication bandwidth. The architecture can be easily implemented in hardware such as in field programmable gate arrays (FPGAs) and application-specific integrated circuits (ASICs).
摘要:
Systems and methods for an inductive powering surface for powering portable devices are described. In one aspect, a powering device includes the inductive powering surface. The inductive powering surface includes multiple primary coils, an impedance auto-match circuit and other control circuits. The impedance auto-match circuit selectively energizes the primary coils to transfer power via inductive coupling to the secondary coil(s) in a portable device. The impedance auto-match circuit is configured to detect voltage and current phase differences over caused by positioning of the portable device on the inductive powering surface. The impedance auto-match circuit calibrates a power factor of the inductive powering surface to transfer an objectively maximized power load via inductive coupling to the portable device.
摘要:
Display components (e.g., liquid crystal displays (LCDs)) are viewable at different viewing angles, for example, by a first user positioned directly in front of the display and by a second user positioned to one side of the display. Many displays present a consistent display across a wide range of viewing angles, but these displays may consume energy and/or compromise user privacy. Presented herein are configurations of backlights for display devices featuring an adjustable viewing angle, such that a user may select a narrower viewing angle in usage scenarios involving power consumption or privacy, and may select a wider viewing angle in usage scenarios involving plentiful power and fewer privacy concerns. Such configurations may include multiple banks of backlight lamps generating backlight at different viewing angles; an electrowetting electrode as an adjustable collimator; adjustable diffusers that may adjustably scatter light passing therethrough; and/or a combination of such adjustable elements.
摘要:
Systems, methods, and devices are described for implementing learning algorithms on data sets. A data set may be partitioned into a plurality of data partitions that may be distributed to two or more processors, such as a graphics processing unit. The data partitions may be processed in parallel by each of the processors to determine local counts associated with the data partitions. The local counts may then be aggregated to form a global count that reflects the local counts for the data set. The partitioning may be performed by a data partition algorithm and the processing and the aggregating may be performed by a parallel collapsed Gibbs sampling (CGS) algorithm and/or a parallel collapsed variational Bayesian (CVB) algorithm. In addition, the CGS and/or the CVB algorithms may be associated with the data partition algorithm and may be parallelized to train a latent Dirichlet allocation model.
摘要:
A user may issue commands to a computing device by moving a pointer within a light field. Sensors may capture light reflected from the moving pointer. A virtual touch engine may analyze the reflected light captured as light portions in a sequence of images by the sensors to issue a command to a computing device in response to the movements. Analyzing the sequence of images may include finding the light portions in the sequence of images, determining a size of the light portions, and determining a location of the light portions.
摘要:
A position determination module is described which receives image information from a video camera. The position determination module identifies at least one reference field in the image information to provide identified reference information. The position determination module then generates position information based on the identified reference information. The position information is associated with a position of the reference field with respect the video camera. In one implementation, the video camera is stationary and the reference field is coupled to a user-manipulated device. In another implementation, the reference field is coupled to a stationary display device and the video camera is coupled to a movable user-manipulated device.