Abstract: A method for generating a target network by performing neural architecture search using optimized search space is provided. The method includes steps of: a computing device (a) if a target data is inputted into the target network, allowing the target network to apply neural network operation to the target data, to generate an estimated search vector; and (b) allowing a loss layer to calculate architecture parameter losses by referring to the estimated search vector and a ground truth search vector, and to perform backpropagation by referring to the architecture parameter losses to update architecture parameter vectors for determining final layer operations among candidate layer operations included in an optimized layer type set corresponding to the optimized search space and wherein the final layer operations are to be performed by neural blocks, within cells of the target network, arranged according to an optimized cell template corresponding to the optimized search space.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for engineering a data analytics platform using machine learning are disclosed. In one aspect, a method includes the actions of receiving data indicating characteristics of data for analysis, analysis techniques to apply to the data, and requirements of users accessing the analyzed data. The actions further include accessing provider information that indicates computing capabilities of a respective data analysis provider, analysis techniques provided by the respective data analysis provider, and real-time data analysis loads of the respective data analysis provider. The actions further include applying the characteristics of the data, the analysis techniques, the requirements of the users, and the provider information, the analysis techniques, and the real-time data analysis loads to a model.

Abstract: One embodiment provides a method comprising extracting natural language content from a piece of communication for a user, generating a representation of the piece of communication based on the natural language content extracted, and utilizing a global deep learning model and a personalized learning model for the user to assign a priority label to the piece of communication based on the representation and user behavioral information associated with recent conversations of the user. Another embodiment provides a method comprising, for each piece of communication of a set of multiple pieces of communication for multiple users, extracting natural language content from the piece communication and generating a representation of the piece of communication based on the natural language extracted, and training a deep learning neural network to predict a degree of priority of a subsequent piece of communication based on each representation generated.

Abstract: In a method for handling a plurality of heuristics for task selection in a genetic algorithm, a task scheduling engine generates a population of tasks associated with an overall objective, identifies multiple jobs associated with an overall objective, compiles the multiple jobs into a genome, and assigns one or more tasks to each job of the multiple jobs. The task scheduling engine also assigns a task heuristic byte defining multiple task heuristics that can be applied to the each job of the genome, randomly assigns a task heuristic from the multiple task heuristics to the each job, and determines a value score for the genome.

Abstract: The technology disclosed determines which field values in a set of unique field values for a particular field in a fielded dataset are anomalous using six similarity measures. A factor vector is generated per similarity measure and combined to form an input matrix. A convolutional neural network processes the input matrix to generate evaluation vectors. A fully-connected network evaluates the evaluation vectors to generate an anomaly scalar for a particular unique field value. Thresholding is applied to anomaly scalar to determine whether the particular unique field value is anomalous.

Abstract: A computing device classifies unclassified observations. A first batch of noise observations is generated. (A) A first batch of unclassified observations is selected. (B) A first batch of classified observations is selected. (C) A discriminator neural network model trained to classify unclassified observations and noise observations is updated with observations that include the first batch of unclassified observations, the first batch of classified observations, and the first batch of noise observations. (D) A discriminator loss value is computed that includes an adversarial loss term computed using a predefined transition matrix. (E) A second batch of unclassified observations is selected. (F) A second batch of noise observations is generated. (G) A generator neural network model trained to generate a fake observation vector for the second batch of noise observations is updated with the second batch of unclassified observations and the second batch of noise observations. (H) (A) to (G) is repeated.

Abstract: A method is disclosed of discriminating detected objects in an area with a vision apparatus. The method includes generating image data of a portion of the area using an imaging device of the object detection device, and processing the image data to classify the image data as an imaged scene type selected from a plurality of scene types stored as scene type data in the memory. The method further includes processing the image data using the object identification data to generate object detection data for each object of the plurality of objects located in the portion of the area, each object detection data having a corresponding scene type of the plurality of scene types obtained from the object identification data, and generating user a sensible output only for the object detection data having a corresponding scene type that is the same as the imaged scene type.

Abstract: In an approach to adjusting user permissions based on mental acuity, one or more computer processors determine whether an individual is within a threshold proximity to a monitored location. In response to determining that an individual is within a threshold proximity to the monitored location, the one or more computer processors identify a required mental acuity for the monitored location. The one or more computer processors determine a current mental acuity for the individual. The one or more computer processors compare the determined mental acuity for the individual with the required mental acuity for the monitored location.

Abstract: In some embodiments, a service platform that facilitates artificial intelligence model and data collection and collection may be provided. Input/output information derived from machine learning models may be obtained via the service platform. The input/output information may indicate (i) first items provided as input to at least one model of the machine learning models, (ii) first prediction outputs derived from the at least one model's processing of the first items, (iii) second items provided as input to at least another model of the machine learning models, (iv) second prediction outputs derived from the at least one other model's processing of the second items, and (v) other inputs and outputs. The input/output information may be provided via the service platform to update a first machine learning model. The first machine learning model may be updated based on the input/output information being provided as input to the first machine learning model.

Abstract: Specification covers new algorithms, methods, and systems for: Artificial Intelligence; the first application of General-AI (versus Specific, Vertical, or Narrow-AI) (as humans can do); addition of reasoning, inference, and cognitive layers/engines to learning module/engine/layer; soft computing; Information Principle; Stratification; Incremental Enlargement Principle; deep-level/detailed recognition, e.g., image recognition (e.g., for action, gesture, emotion, expression, biometrics, fingerprint, tilted or partial-face, OCR, relationship, position, pattern, and object); Big Data analytics; machine learning; crowd-sourcing; classification; clustering; SVM; similarity measures; Enhanced Boltzmann Machines; Enhanced Convolutional Neural Networks; optimization; search engine; ranking; semantic web; context analysis; question-answering system; soft, fuzzy, or un-sharp boundaries/impreciseness/ambiguities/fuzziness in class or set, e.g.

Abstract: Techniques are provided for imputation in computer-based reasoning systems. The techniques include performing the following until there are no more cases in a computer-based reasoning model with missing fields for which imputation is desired: determining which cases have fields to impute (e.g., missing fields) in the computer-based reasoning model and determining conviction scores and/or imputation order information for the cases that have fields to impute. The techniques proceed by determining for which cases to impute data and, for each of the determined one or more cases with missing fields to impute data is imputed for the missing field, and the case is modified with the imputed data. Control of a system is then caused using the updated computer-based reasoning model.

Abstract: A neuromorphic computing system includes a synapse array, a switching circuit, a sensing circuit and a processing circuit. The synapse array includes row lines, column lines and synapses. The processing circuit is coupled to the switching circuit and the sensing circuit and is configured to connect a particular column line in the column lines to the first terminal by using the switching circuit, obtain a first voltage value from the particular column line by using the sensing circuit when the particular line is connected to the first terminal, connect the particular column line to the second terminal by using the switching circuit, obtain a second voltage value from the particular column line by using the sensing circuit when the particular line is connected to the second terminal, and estimate a sum-of-product sensing value according to a voltage difference between the first voltage value and the second voltage value.

Abstract: An architecture for an explainable neural network may implement a number of layers to produce an output. The input layer may be processed by both a conditional network and a prediction network. The conditional network may include a conditional layer, an aggregation layer, and a switch output layer. The prediction network may include a feature generation and transformation layer, a fit layer, and a value output layer. The results of the switch output layer and value output layer may be combined to produce the final output layer. A number of different possible activation functions may be applied to the final output layer depending on the application. The explainable neural network may be implementable using both general purpose computing hardware and also application specific circuitry including optimized hardware only implementations. Various embodiments of XNNs are described that extend the functionality to different application areas and industries.

Abstract: Aspects of the present disclosure involve systems, methods, devices, and the like for generating compact tree representations applicable to machine learning. In one embodiment, a system is introduced that can retrieve a decision tree structure to generate a compact tree representation model. The compact tree representation model may come in the form of a matrix design to maintain the relationships expressed by the decision tree structure.

Abstract: A data processing apparatus may utilize an artificial neuron network configured to reduce dimensionality of input data using a sparse transformation configured using receptive field structure of network units. Output of the network may be analyzed for temporally persistency that is characterized by similarity matrix. Elements of the matrix may be incremented when present activity unit activity at a preceding frame. The similarity matrix may be partitioned based on a distance measure for a given element of the matrix and its closest neighbors. Stability of learning of temporally proximal patterns may be greatly improved as the similarity matrix is learned independently of the partitioning operation. Partitioning of the similarity matrix using the methodology of the disclosure may be performed online, e.g., contemporaneously with the encoding and/or similarity matrix construction, thereby enabling learning of new features in the input data.

Abstract: Methods and apparatus related to determining a semantically diverse subset of candidate responses to provide for initial presentation to a user as suggestions for inclusion in a reply to an electronic communication. Some of those implementations determine the semantically diverse subset of candidate responses based on generating, over a neural network response encoder model, embeddings that are each based on one of the plurality of the candidate responses. The embedding based on a given candidate response may be compared to embedding(s) of candidate response(s) already selected for the subset, and the given candidate response added to the subset only if the comparing indicates a difference criterion is satisfied.

Abstract: A virtual assistant ecosystem is presented. One can instantiate or construct a customized virtual assistant when needed by capturing a digital representation of one or more objects. A virtual assistant engine analyzes the digital representation to determine the nature or type of the objects present. The engine further obtains attributes for a desirable assistant based on the type of objects. Once the attributes are compiled the engine can then create the specific type of assistant required by the circumstances.

Abstract: A method for searching a new material includes: performing a learning on a material model, which is modeled based on a known material; determining a candidate material by inputting a targeted physical property to a result of the learning; and determining the new material from the candidate material.

Abstract: Techniques for identifying weaknesses in a probabilistic model such as an artificial neural network using an iterative process are disclosed. A seed file may be obtained and variant files generated therefrom. The variant files may be evaluated for their fitness, based upon the ability of the variant files to cause the probabilistic model to fail. The fittest variants, which may refer to those variants that are most successful in causing the model to fail, may be selected. From these selected variants, a next generation of variant files may be created. The next generation of variant files may be evaluated for their fitness. At each step of fitness evaluation or at the end of the iterative process, a map of the fittest variants may be generated to identify hotspots. These hotspots may reveal segments of code or a file that are problematic for the model, which can be used to improve the model.

Abstract: Mechanisms for training a Question and Answer (QA) system are provided. The QA system receives a training question for processing by the QA system and processes the training question to generate an answer to the training question, from a portion of content. The QA system identifies a repeatable pattern of content present in the portion of content in association with the answer to the training question. The QA system applies the repeatable pattern of content to other portions of content to generate at least one additional training question and at least one additional entry in a ground truth data structure to thereby expand a set of training questions and expand the ground truth data structure. The QA system is then trained using the expanded set of training questions and expanded ground truth data structure.