Abstract: Disclosed herein are an apparatus and method for detecting a keypoint based on deep learning robust to scale changes based on information change across receptive fields. The apparatus for detecting a keypoint based on deep learning robust to scale changes based on information change across receptive fields includes a feature extractor for extracting a feature from an input image based on a pre-trained deep learning neural network, an information accumulation pyramid module for outputting, from the feature, at least two filter responses corresponding to receptive fields having different scales, an information change detection module for calculating an information change between the at least two filter responses, a keypoint detection module for creating a score map having a keypoint probability of each pixel based on the information change, and a continuous scale estimation module for estimating a scale of a receptive field having a biggest information change for each pixel.

Abstract: Disclosed are a multi-task training technique and resulting model for detecting distracted driving. In one embodiment, a method is disclosed comprising inputting a plurality of labeled examples into a multi-task network, the multi-task network comprising: a backbone network, the backbone network generating one or more feature vectors corresponding to each of the labeled examples, and a plurality of prediction heads coupled to the backbone network; minimizing a joint loss based on outputs of the plurality of prediction heads, the minimizing the joint loss causing a change in parameters of the backbone network; and storing a distraction classification model after minimizing the joint loss, the distraction classification model comprising the parameters of the backbone network and parameters of at least one of the prediction heads.

Abstract: Methods and systems for analyzing, ordering and presenting item listings are described. In one example embodiment, a search query is processed to identify item listings satisfying the search query. Then, for each item listing that satisfies the search query, a ranking score is derived and assigned to the item listing. The ranking score is based in part on a relevance score, a listing quality score and a business rules score (or, adjustment factor). Finally, the item listings are ordered, based on their corresponding ranking score, and presented in order in a search results page.

Abstract: Disclosed are a multi-task training technique and resulting model for detecting distracted driving. In one embodiment, a method is disclosed comprising inputting a plurality of labeled examples into a multi-task network, the multi-task network comprising: a backbone network, the backbone network generating one or more feature vectors corresponding to each of the labeled examples, and a plurality of prediction heads coupled to the backbone network; minimizing a joint loss based on outputs of the plurality of prediction heads, the minimizing the joint loss causing a change in parameters of the backbone network; and storing a distraction classification model after minimizing the joint loss, the distraction classification model comprising the parameters of the backbone network and parameters of at least one of the prediction heads.

Abstract: First and second instruction storage are coupled with a fetch unit including sets of fetch circuitry each spanning a plurality of pipeline stages. A first set of fetch circuitry is to initiates a fetch operation for a block of instructions, and has an indication to read the block of instructions from the second instruction storage. The first set retains the fetch operation for the block of instructions at a pipeline stage of the plurality, for one or more cycles, until a hazard corresponding to the pipeline stage of the first set of fetch circuitry has been removed. The first set stores the block of instructions from the second instruction storage to the first instruction storage, during the one or more cycles. The first set reads the block of instructions from the first instruction storage, for the fetch operation, once the hazard has been removed.

Abstract: At least one instruction storage coupled with a fetch unit including sets of fetch circuitry each having a same plurality of pipeline stages. The sets of fetch circuitry perform fetch operations to fetch blocks of instructions from the at least one instruction storage. Stall circuitry, in response to an indication of a hazard for a given pipeline stage of a first set of fetch circuitry, retains a fetch operation for a first block of instructions at the given pipeline stage, and zero or more fetch operations for zero or more corresponding blocks of instructions at zero or more preceding pipeline stages of the first set of fetch circuitry, until the hazard has been removed. The stall circuitry advances a fetch operation for a second block of instructions from the given pipeline stage of a second set of fetch circuitry, during an initial cycle of the one or more cycles.

Abstract: Disclosed herein is an image deep learning model training method. The method includes sampling a twin negative comprising a first negative sample and a second negative sample by selecting the first negative sample with a highest similarity out of an anchor sample and a positive sample constituting a matching pair in each class and by selecting the second negative sample with a highest similarity to the first negative sample, and training the samples to minimize a loss of a loss function in each class by utilizing the anchor sample, the positive sample, the first and second negative samples for each class. The first negative sample is selected in a different class from a class comprising the matching pair, and the second negative sample is selected in a different class from classes comprising the matching pair and the first negative sample.

Abstract: Disclosed are a multi-task training technique and resulting model for detecting distracted driving. In one embodiment, a method is disclosed comprising inputting a plurality of labeled examples into a multi-task network, the multi-task network comprising: a backbone network, the backbone network generating one or more feature vectors corresponding to each of the labeled examples, and a plurality of prediction heads coupled to the backbone network; minimizing a joint loss based on outputs of the plurality of prediction heads, the minimizing the joint loss causing a change in parameters of the backbone network; and storing a distraction classification model after minimizing the joint loss, the distraction classification model comprising the parameters of the backbone network and parameters of at least one of the prediction heads.

Abstract: Disclosed herein are an apparatus and method for detecting a keypoint based on deep learning robust to scale changes based on information change across receptive fields. The apparatus for detecting a keypoint based on deep learning robust to scale changes based on information change across receptive fields includes a feature extractor for extracting a feature from an input image based on a pre-trained deep learning neural network, an information accumulation pyramid module for outputting, from the feature, at least two filter responses corresponding to receptive fields having different scales, an information change detection module for calculating an information change between the at least two filter responses, a keypoint detection module for creating a score map having a keypoint probability of each pixel based on the information change, and a continuous scale estimation module for estimating a scale of a receptive field having a biggest information change for each pixel.

Abstract: Disclosed are a multi-task training technique and resulting model for detecting distracted driving. In one embodiment, a method is disclosed comprising inputting a plurality of labeled examples into a multi-task network, the multi-task network comprising: a backbone network, the backbone network generating one or more feature vectors corresponding to each of the labeled examples, and a plurality of prediction heads coupled to the backbone network; minimizing a joint loss based on outputs of the plurality of prediction heads, the minimizing the joint loss causing a change in parameters of the backbone network; and storing a distraction classification model after minimizing the joint loss, the distraction classification model comprising the parameters of the backbone network and parameters of at least one of the prediction heads.

Abstract: Disclosed are a multi-task training technique and resulting model for detecting distracted driving. In one embodiment, a method is disclosed comprising inputting a plurality of labeled examples into a multi-task network, the multi-task network comprising: a backbone network, the backbone network generating one or more feature vectors corresponding to each of the labeled examples, and a plurality of prediction heads coupled to the backbone network; minimizing a joint loss based on outputs of the plurality of prediction heads, the minimizing the joint loss causing a change in parameters of the backbone network; and storing a distraction classification model after minimizing the joint loss, the distraction classification model comprising the parameters of the backbone network and parameters of at least one of the prediction heads.

Abstract: A method of improving a ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCo) to have a higher protein score is disclosed. The method includes the steps of: making a modified RbcL of the RuBisCo, by, on an RbcL unit of the RuBisCo, either substituting Met for Leu, Phe, Val, or Ile or combinations thereof; substituting Lys for Arg, Thr, or His or combinations thereof; or both of these substitutions. The modified RbcL consequently modifies the RuBisCo and is added to a biomass host where it is stable for homologous recombination. Plastid and nucleus integration was observed. Example RbcL sequences are disclosed with the desirable substitutions. The improved RuBisCo can be used as an improved proteinaceous food source for humans and animals.

Abstract: An embodiment includes determining if a new incident report of a new incident matches any resolved incident reports associated with resolved incidents. The embodiment performs a first classification operation on the new incident report to determine if the new incident report is likely to be similar to any resolved incident reports associated with resolved incidents. The embodiment also performs a second classification operation on the new incident report to generate a ranked list of changes that are likely to be similar to the new incident report. The embodiment outputs the ranked list of changes to an incident manager for evaluation, then receives an input representative of a selected change from among the ranked list of changes responsible for causing the new incident. The embodiment revises the new incident report to include a reference to the selected change.

Abstract: Methods and systems enable access to listing performance data associated with a search engine. A request for listing performance data for an item is received from a client computing system. The request is associated with a keyword and/or a category. The request is processed to retrieve the listing performance data corresponding to the item, and a value is determined that indicates a strength of the item in a search results listing having a plurality of search results based on a search using the keyword and/or the category. The listing performance data and value are provided for presentation.

Abstract: Examples described herein provide a computer-implemented method that includes aggregating data. The method further includes filtering the data to eliminate objects known not to be suitable for additive manufacturing. The method further includes performing machine learning on the data to identify objects suitable for additive manufacturing. The method further includes generating, using additive manufacturing, at least one of the objects identified as being suitable for additive manufacturing.

Abstract: Methods and systems for analyzing, ordering and presenting item listings are described. In one example embodiment, a search query is processed to identify item listings satisfying the search query. Then, for each item listing that satisfies the search query, a ranking score is derived and assigned to the item listing. The ranking score is based in part on a relevance score, a listing quality score and a business rules score (or, adjustment factor). Finally, the item listings are ordered, based on their corresponding ranking score, and presented in order in a search results page.

Abstract: Methods and systems for analyzing, ordering and presenting item listings are described. In one example embodiment, a search query is processed to identify item listings satisfying the search query. Then, for each item listing that satisfies the search query, a ranking score is derived and assigned to the item listing. The ranking score is based in part on a relevance score, a listing quality score and a business rules score (or, adjustment factor). Finally, the item listings are ordered, based on their corresponding ranking score, and presented in order in a search results page.

Abstract: An integrated circuit (IC) includes multiple interconnected driver cells enabled/disabled based on a first set of control signals. The multiple circuit cells are arranged to define a first aggregate enabled/disabled configuration exhibiting a first aggregated delay. The first aggregated delay is based on the individual enabled/disabled states of the circuit cells. Timing circuitry evaluates the first aggregate delay with respect to a circuit design constraint, and selectively generates a second set of control signals to configure the multiple circuit cells to define a second aggregate enabled/disabled configuration having a second aggregate delay different than the first aggregate delay.

Abstract: Disclosed herein is an image deep learning model training method. The method includes sampling a twin negative comprising a first negative sample and a second negative sample by selecting the first negative sample with a highest similarity out of an anchor sample and a positive sample constituting a matching pair in each class and by selecting the second negative sample with a highest similarity to the first negative sample, and training the samples to minimize a loss of a loss function in each class by utilizing the anchor sample, the positive sample, the first and second negative samples for each class. The first negative sample is selected in a different class from a class comprising the matching pair, and the second negative sample is selected in a different class from classes comprising the matching pair and the first negative sample.

Abstract: An integrated circuit (IC) includes multiple interconnected driver cells enabled/disabled based on a first set of control signals. The multiple circuit cells are arranged to define a first aggregate enabled/disabled configuration exhibiting a first aggregated delay. The first aggregated delay is based on the individual enabled/disabled states of the circuit cells. Timing circuitry evaluates the first aggregate delay with respect to a circuit design constraint, and selectively generates a second set of control signals to configure the multiple circuit cells to define a second aggregate enabled/disabled configuration having a second aggregate delay different than the first aggregate delay.