Abstract: Systems and methods are described for time-sharing interactions using a shared artificial intelligence personality (AIP) incorporated within multiple human interaction entities (HIEs). An AIP is an understanding construct that may control a variety of communication experiences to support a sense of ongoing social connectedness. An AIP may be instantiated within two or more HIEs that interact with humans in a human, cartoon or pet-like manner. HIEs may include robots, robotic pets, toys, simple-to-use devices, and graphical user interfaces. The AIP may be periodically updated based on human interactions sensed by the HIEs as well as knowledge of historical and ongoing events. The systems may provide two or more users with intuitive machine companions that exhibit an expert knowledge base and a familiar, cumulative personality.

Abstract: A method for determining a final architecture for a neural network to perform a particular machine learning task is described.

Abstract: In one embodiment, a set of training data consisting of inliers may be obtained. A supervised classification model may be trained using the set of training data to identify outliers. The supervised classification model may be applied to generate an anomaly score for a data point. It may be determined whether the data point is an outlier based, at least in part, upon the anomaly score.

Abstract: The present disclosure generally relates to analysis and detection of human emotions, and more particularly, to a method and system for detection of human emotion based on an individual's typing behavior. In one embodiment, the method comprises of receiving typing data of the individual for a specified amount of time, processing the received data to extract hand speed and key-press duration for the individual and analyzing the hand speed and key-press duration to detect probable emotional states the individual has experienced for the specified amount of time.

Abstract: System and method for analyzing driver behavior based on telematics data are disclosed. In an example, a probability of a user driving a vehicle is computed and a risk score is generated to develop at least one driver profile based on the probability. Further, routes taken by said user driving said vehicle are clustered to generate enhanced driver profile and using the clustered output to develop dynamic intelligent contexts for each said route and adding contextual intelligence messages to customize said risk score. Furthermore, the routes taken by the said user in real time are predicted. In addition, a missing route is identified through imputation of missed routes to compute annualized mileage, and a missing distance is imputed in an analysis of at least one trip of the driver in the vehicle. Also, independent trips are stitched based on at least one recommendation from an analytics engine.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for reinforcement learning. The embodiments described herein apply meta-learning (and in particular, meta-gradient reinforcement learning) to learn an optimum return function G so that the training of the system is improved. This provides a more effective and efficient means of training a reinforcement learning system as the system is able to converge on an optimum set of one or more policy parameters ? more quickly by training the return function G as it goes. In particular, the return function G is made dependent on the one or more policy parameters ? and a meta-objective function J? is used that is differentiated with respect to the one or more return parameters ? to improve the training of the return function G.

Abstract: A neural network training method based on training data, includes receiving training data including sequential data, and selecting a reference hidden node from hidden nodes in a neural network. The method further includes training the neural network based on remaining hidden nodes obtained by excluding the reference hidden node from the hidden nodes, and based on the training data, the remaining hidden nodes being connected with hidden nodes in a different time interval, and a connection between the reference hidden node and the hidden nodes in the different time interval being ignored.

Abstract: An inference is made regarding whether or not an upcoming day is going to be a busy day for a user. One or more different user-specific event parameters are utilized to compute a user busyness score for the upcoming day, where these parameters are based in part on a history of events for the user and their past behavior. Then, whenever the user busyness score for the upcoming day is greater than a busy day threshold, it is inferred that the upcoming day is going to be a busy day for the user. Whenever the user busyness score for the upcoming day is less than a quiet day threshold, it is inferred that the upcoming day is going to be a quiet day for the user.

Abstract: Systems and methods to analyze decision-making. A processor accesses at least one storage medium storing a high-fidelity recording, the high-fidelity recording includes a stimulus event and one or more rules associated with the stimulus event. The processor presents the high-fidelity recording through a presentation device. The processor receives neural metrics from a neural sensor while the high-fidelity recording is being presented. The processor applies rules to generate reference metrics associated with the stimulus event. The processor modifies elements of the high-fidelity recording based on the reference metrics to distinguish the stimulus event within the high fidelity recording. The processor presents the modified recording by the presentation device.

Abstract: Memory facilitation using directed acyclic graphs is described, for example, where a plurality of directed acyclic graphs are trained for gesture recognition from human skeletal data, or to estimate human body joint positions from depth images for gesture detection. In various examples directed acyclic graphs are grown during training using a training objective which takes into account both connection patterns between nodes and split function parameter values. For example, a layer of child nodes is grown and connected to a parent layer of nodes using an initialization strategy. In examples, various local search processes are used to find good combinations of connection patterns and split function parameters.

Abstract: Data is classified using automatically selected hyperparameter values. (A) A first loss value is determined based on a converged classification matrix. (B) Each observation vector is assigned to a cluster using a clustering algorithm based on the converged classification matrix. (C) A predefined number of observation vectors is selected from each cluster. D) Classified observation vectors and unclassified observation vectors are updated based on the selections in (C) and (A) is repeated. (E) An entropy loss value is determined, wherein (A) to (E) are repeated for a plurality of different values of a kernel parameter value and a batch size value. (F) A second loss value is determined based on the converged classification matrix, a label matrix defined from the converged classification matrix, and a weight value. (L) (A) to (F) are repeated with a plurality of different values of the weight value until convergence is satisfied.

Abstract: A method and system for linking a multimedia data element (MMDE) and a webpage are provided. The method includes receiving a MMDE from a source; generating a signature representative of the MMDE using a plurality of computational cores; matching the generated signature with a plurality of signatures stored in a database to find at least one matching signature, wherein at least one of the stored signatures has at least one corresponding universal resource locator (URL) of a web page stored therein as metadata of the at least one of the stored signatures; and providing to the source at least a URL that is a metadata of a matched signature upon determination of a match between the generated signature and at least one of the plurality of signatures stored in the database.

Abstract: A reconfigurable neural network circuit is provided. The reconfigurable neural network circuit comprises an electronic synapse array including multiple synapses interconnecting a plurality of digital electronic neurons. Each neuron comprises an integrator that integrates input spikes and generates a signal when the integrated inputs exceed a threshold. The circuit further comprises a control module for reconfiguring the synapse array. The control module comprises a global final state machine that controls timing for operation of the circuit, and a priority encoder that allows spiking neurons to sequentially access the synapse array.

Abstract: The present invention relates to artificial neural network (ANN), for example, convolutional neural network (CNN). In particular, the present invention relates to how to implement and optimize a convolutional neural network based on an embedded FPGA. Specifically, it proposes a CPU+FPGA heterogeneous architecture to accelerate ANNs.

Abstract: The present disclosure relates to a method and attention neural network for automatically learning embeddings for various latent aspects of textual claims and documents performed in an attention neural network comprising one or more latent aspect models for guiding an attention mechanism of the neural network, wherein the method comprises the steps of inserting a claim document pair, in each of the latent aspect models and a latent aspect vector to select significant sentences to form document representations for each respective latent aspect of the latent aspect vector, concatenating the document representations to establish an overall document representation, calculating a class probability distribution by means of the overall document representation, and classifying the claim of document as true or false using the class probability distribution.

Abstract: A computer trains a neural network model. (B) A neural network is executed to compute a post-iteration gradient vector and a current iteration weight vector. (C) A search direction vector is computed using a Hessian approximation matrix and the post-iteration gradient vector. (D) A step size value is initialized. (E) An objective function value is computed that indicates an error measure of the executed neural network. (F) When the computed objective function value is greater than an upper bound value, the step size value is updated using a predefined backtracking factor value. The upper bound value is computed as a sliding average of a predefined upper bound updating interval value number of previous upper bound values. (G) (E) and (F) are repeated until the computed objective function value is not greater than the upper bound value. (H) An updated weight vector is computed to describe a trained neural network model.

Abstract: Methods, systems, and computer-readable storage media for providing an insight provider including a logic component and a configuration component, the logic component including a domain-specific model, the configuration component including one or more parameter values for processing data using the domain-specific model, receiving a set of assets including data indicative of one or more assets, retrieving asset data associated with at least one asset of the first set of assets, the asset data including OT data and IT data, the OT data being provided from one or more networked devices, the IT data being provided from one or more enterprise systems, and processing the OT data and the IT data using the domain-specific model of the logic component to provide a result set, the result set including one or more of a second set of assets and enriched data.

Abstract: Systems and methods are described for sharing an artificial intelligence personality (AIP) among multiple human interaction entities (HIEs). An AIP is an understanding construct that interacts with one or more humans in a human- or pet-like manner, implementing a variety of communication experiences to support a sense of ongoing social connectedness. HIEs may include robots, robotic pets, toys, and avatars. The system may be implemented using two or more HIEs and, optionally, one or more remote and/or distributed processors to compute AIPs and sequence telecommunications. AIPs are periodically updated primarily based on human interactions sensed by the two or more HIEs. HIEs may continue to operate without interruption in the presence of significant telecommunications delays. The systems may provide two or more users with intuitive machine companions that exhibit an integrated knowledge base and personality cumulatively acquired from all users.

Abstract: Customer journey prediction and resolution is accomplished via a predictive model in which each user is mapped onto all available user journey information corresponding to a specific business. The predictive model is analyzed to understand the characteristics, preferences, and lowest effort resolution for the user related to the services that are subscribed to by the user. The predictive model is analyzed to predict the service or collection of services for each user. Embodiments interact with, provide and receive information from, and react to and/or deliver action to the customer across channels and across services. All customer and system behavior, data, and action is tracked and coordinated and leveraged for continuous feedback and performance improvement.

Abstract: An approach for predictively scoring test case results in real-time. Test case results associated with a test run are received by a software testing environment. Using predictive statistical models, test case results and attribute relationships are matched against model rules and test case history. A statistical correlation and confidence parameter provide the ability to generate test case relationships for predicting the outcome of other test cases in the test run. The test case relationships are transformed into scoring results and output for the further processing.