Abstract: An electronic control device is an electronic control device that stores a neural network including plural neurons and a connection information group including plural pieces of connection information that associate the neurons with each other. The neural network is a neural network including a third connection information group created through deletion, by a pruning section, of at least one piece of connection information from a first connection information group for deciding an output value when an input value is given on the basis of the first connection information group and a second connection information group that is plural pieces of connection information having the degree of influence on predetermined data given as the input value, the degree of influence exceeding a predetermined value.

Abstract: Systems, methods, and apparatuses for providing dynamic, prioritized spectrum utilization management. The system includes at least one monitoring sensor, at least one data analysis engine, at least one application, a semantic engine, a programmable rules and policy editor, a tip and cue server, and/or a control panel. The tip and cue server is operable utilize the environmental awareness from the data processed by the at least one data analysis engine in combination with additional information to create actionable data.

Abstract: To provide a scheme capable of easily generating knowledge data containing an enormous number of assumed input sentences (input sentences assumed to be inputted by a user) in an information providing system realizing a conversation with the user by outputting responses corresponding to inputs by the user to a terminal of the user. An information providing system develops sets of assumed input sentences and response sentences with respect to respective data stored in intermediate data based on a set of assumed input sentences and response sentences defined in template data in accordance with an instruction of a client or the like to generate knowledge data.

Abstract: Provided are a latency prediction method and a computing device for the same. The latency prediction method includes receiving a deep learning model and predicting on-device latency of the received deep learning model using a latency predictor which is trained on the basis of a latency lookup table. The latency lookup table includes information on single neural network layers and latency information of the single neural network layers on an edge device.

Abstract: Computer-implemented systems find methods for synthesis of concept definitions and concept relationships from a domain of data, utilizing different semantic processing protocols such formal concept analysis and faceted classification synthesis from existing domain concepts that have a confidence gradient built into them. A cognitive or an input agent provides an input of an active concept which is matched against existing domain concepts. The resultant pool of relevant domain concepts is then used to derive virtual concept definitions using a semantic processing protocol. The derivation is then overlaid with a concept of relative proximity of an attribute from another within an attribute set. An additional layer of coherence is given by the relative proximity measure. The end result is a pool of related virtual concept definitions in a tree structure.

Abstract: Techniques for an iterative singular spectrum analysis are provided. In one technique, a first analysis, of time series data, is performed that results in a first reconstructed version of the time series data. The first analysis, of the time series data and a portion of the first reconstructed version, is then performed that results in a second reconstructed version of the time series data. Based on a termination condition, it is determined whether to perform the first analysis relative to a portion of a third reconstructed version of the time series data. A second analysis, of the time series data and a portion of a fourth reconstructed version of the time series data, is performed that results in a fifth reconstructed version of the time series data. The second analysis is different than the first analysis. A difference between the time series data and the fifth reconstructed version data is computed.

Abstract: Techniques are provided for sparse ensembling of unsupervised machine learning models. In an embodiment, the proposed architecture is composed of multiple unsupervised machine learning models that each produce a score as output and a gating network that analyzes the inputs and outputs of the unsupervised machine learning models to select an optimal ensemble of unsupervised machine learning models. The gating network is trained to choose a minimal number of the multiple unsupervised machine learning models whose scores are combined to create a final score that matches or closely resembles a final score that is computed using all the scores of the multiple unsupervised machine learning models.

Abstract: Methods and systems for analyzing machine-learned classifiers are disclosed herein. The method can include inputting a data item for processing by a machine-learned classifier model and receiving a plurality of confidence scores for a plurality of respective classes, the plurality of confidence scores having been generated by the machine-learned classifier model based on the data item. The method can also include determining a distance in dependence on a highest confidence score that is generated for the data item, and causing display of a class distribution diagram, where the class distribution diagram can illustrate a graphical representation corresponding to the data item located at said distance between the graphical representation of a first class and the graphical representation of a second class.

Abstract: A control system includes: a generation unit that generates a dataset for each unit section; a feature extraction unit that generates feature quantity data on the basis of the dataset; and a score calculation unit that calculates a score indicating a degree of deviation of the feature quantity data from learning data, by referring to the learning data. The feature quantity data and the score are output as compression results of the dataset. The control system includes a restoration unit that selects pattern data corresponding to a class determined according to the score contained in the compression result, and after adjusting the pattern data using the feature quantity data contained in the compression results, restores the pattern data as a temporal change in the dataset corresponding to the compression results.

Abstract: Systems, methods, and apparatuses for providing dynamic, prioritized spectrum utilization management. The system includes at least one monitoring sensor, at least one data analysis engine, at least one application, a semantic engine, a programmable rules and policy editor, a tip and cue server, and/or a control panel. The tip and cue server is operable utilize the environmental awareness from the data processed by the at least one data analysis engine in combination with additional information to create actionable data.

Abstract: Embodiments of a method are disclosed. The method includes performing distributed deep learning training on multiple batches of training data using corresponding learners. Additionally, the method includes determining training times wherein the learners perform the distributed deep learning training on the batches of training data. The method also includes modifying a processing aspect of the straggler to reduce a future training time of the straggler for performing the distributed deep learning training on a new batch of training data in response to identifying a straggler of the learners by a centralized control.

Abstract: A copy of a model comprising a plurality of trees is received, as is a copy of training set data comprising a plurality of training set examples. For each tree included in the plurality of trees, the training set data is used to determine which training set examples are classified as a given leaf. A blame forest is generated at least in part by mapping each training set item to the respective leaves at which it arrives.

Abstract: Systems, methods, and computer program products for predicting user preference of items in an image process image data associated with a single image with a first branch of a neural network to produce an image embedding, the single image including a set of multiple items; process a user identifier of a user with a second branch of the neural network to produce a user embedding; concatenate the image embedding with the user embedding to produce a concatenated embedding; process the concatenated embedding with the neural network to produce a joint embedding; and generate a user preference score for the set of multiple items from the neural network based on the joint embedding, the user preference score including a prediction of whether the user prefers the set of multiple items.

Abstract: In order to facilitate the entity resolution and entity activity tracking and indexing, systems and methods include receiving first source records from a first database and second source records from a record database. A candidate set of second source records is determined by a heuristic search in the set of second source records. A candidate pair feature vector associated with each candidate pair of first and second source records is generated. An entity matching machine learning model predicts matching first source records for each candidate second source record based on the respective candidate pair feature vector. An aggregate quantity associated with the matching first source records is aggregated from a quantity associated with each first source record, and a quantity index for each candidate second source record is determined based the aggregate quantities. Each quantity index is displayed to a user.

Abstract: A system and method may partition and/or execute a NN, by, for a graph including nodes and hyper edges, each node representing a data item in the NN and each hyper edge representing an operation in the NN, identifying a deep tensor column comprising a subset of the nodes and a subset of the hyper edges, such that the operations in the deep tensor column, when executed, use only data which fits within a preselected cache.

Abstract: A method for accelerating machine learning on a computing device is described. The method includes hosting a neural network in a first inference accelerator and a second inference accelerator. The neural network split between the first inference accelerator and the second inference accelerator. The method also includes routing intermediate inference request results directly between the first inference accelerator and the second inference accelerator. The method further includes generating a final inference request result from the intermediate inference request results.

Abstract: Systems and methods are described for facilitating operation of a plurality of computing devices. Data indicative of enumerated resources of a computing device is collected. The data is collected without dependency on write permissions to a file system of the one computing device. A condition of the computing device is determined based on historical data associated with enumerated resources of other computing devices. The identified condition can be updated as updated historical data becomes available. A communication to the computing device may be sent based on the identified condition.

Abstract: Embodiments are directed to a composite binary decomposition network. An embodiment of a computer-readable storage medium includes executable computer program instructions for transforming a pre-trained first neural network into a binary neural network by processing layers of the first neural network in a composite binary decomposition process, where the first neural network having floating point values representing weights of various layers of the first neural network. The composite binary decomposition process includes a composite operation to expand real matrices or tensors into a plurality of binary matrices or tensors, and a decompose operation to decompose one or more binary matrices or tensors of the plurality of binary matrices or tensors into multiple lower rank binary matrices or tensors.

Abstract: The present disclosure is directed to an automated neural architecture search approach for designing new neural network architectures such as, for example, resource-constrained mobile CNN models. In particular, the present disclosure provides systems and methods to perform neural architecture search using a novel factorized hierarchical search space that permits layer diversity throughout the network, thereby striking the right balance between flexibility and search space size. The resulting neural architectures are able to be run relatively faster and using relatively fewer computing resources (e.g., less processing power, less memory usage, less power consumption, etc.), all while remaining competitive with or even exceeding the performance (e.g., accuracy) of current state-of-the-art mobile-optimized models.

Abstract: Systems and methods are described for training a machine learning model using intelligently selected multiclass vectors. According to an embodiment, a set of un-labeled feature vectors are received. The set of feature vectors are grouped into clusters within a vector space having fewer dimensions than the first set of feature vectors by applying a homomorphic dimensionality reduction algorithm to the set of feature vectors and performing centroid-based clustering. An optimal set of clusters among the clusters is identified by performing a convex optimization process on the clusters. Vector labeling is minimized by selecting ground truth representative vectors including a representative vector from each cluster of the optimal set of clusters. A set of labeled feature vectors is created based on labels received from an oracle for each of the representative vectors. A machine-learning model is trained for multiclass classification based on the set of labeled feature vectors.