Abstract: Systems and methods are disclosed for operating a machine, by receiving training data from one or more sensors; training a machine learning module with the training data by: partitioning a data matrix into smaller submatrices to process in parallel and optimized for each processing node; for each submatrix, performing a greedy search for rank-one solutions; using alternating direction method of multipliers (ADMM) to ensure consistency over different data blocks; and controlling one or more actuators using live data and the learned module during operation.

Abstract: A processing device determines a plurality of visual concepts for visual data based on at least one of visual entities in the visual data or feature-level attributes in the visual data, wherein the visual entities are based on the feature-level attributes, and wherein each of the plurality of visual concepts comprises a subject visual entity related to an object visual entity by a predicate. The processing device further determines one or more visual semantics for the visual data based on the plurality of visual concepts, wherein the one or more visual semantics define relationships between the plurality of visual concepts.

Abstract: An information processing method and apparatus, the method including: training a Deep Neural Network (DNN) by using an evaluation object seed, an evaluation term seed and an evaluation relationship seed (101); at a first input layer, connecting vectors corresponding to a candidate evaluation object, a candidate evaluation term and a candidate evaluation relationship to obtain a first input vector (102); at a first hidden layer, compressing the first input vector to obtain a first middle vector, and at a first output layer, decoding the first middle vector to obtain a first output vector (103); and determining a first output vector whose decoding error value is less than a decoding error value threshold, and determining a candidate evaluation object, a candidate evaluation term and a candidate evaluation relationship corresponding to the determined first output vector as first opinion information (104).

Abstract: The disclosure is directed to evaluating feature vectors using decision trees. Typically, the number of feature vectors and the number of decision trees are very high, which prevents loading them into a processor cache. The feature vectors are evaluated by processing the feature vectors across a disjoint subset of trees repeatedly. After loading the feature vectors into the cache, they are evaluated across a first subset of trees, then across a second subset of trees and so on. If the values based on the first and second subsets satisfy a specified criterion, further evaluation of the feature vectors across the remaining of the decision trees is terminated, thereby minimizing the number of trees evaluated and therefore, consumption of computing resources.

Abstract: A system for creating and using a universal tag to gather consumer data from a web site for the purposes of targeted advertising is provided. The universal tag system has two main subsystems. The first subsystem is a configuration system that is used to define the consumer data to be collected from the website and to define taxonomy and transformation rules to be applied to the collected consumer data. The second subsystem is a runtime system that runs a universal tag client-side script, which is triggered when a consumer lands on a webpage of the website, for collecting the defined consumer data. The runtime system then applies the transformation rules to the collected data and updates a user profile corresponding to the consumer with the transformed data. As well, the runtime system applies the taxonomy rules to the collected data and categorizes the consumer for the purposes of subsequent targeted advertising.

Abstract: A method, article comprising machine-readable instructions and apparatus that processes data systems for encoding, decoding, pattern recognition/matching and data generation is disclosed. State subsets of a data system are identified for the efficient processing of data based, at least in part, on the data system's systemic characteristics.

Abstract: Embodiments of the invention relate to a globally asynchronous and locally synchronous neuromorphic network. One embodiment comprises generating a synchronization signal that is distributed to a plurality of neural core circuits. In response to the synchronization signal, in at least one core circuit, incoming spike events maintained by said at least one core circuit are processed to generate an outgoing spike event. Spike events are asynchronously communicated between the core circuits via a routing fabric comprising multiple asynchronous routers.

Abstract: Data sets for a three-stage predictor can be automatically determined. For example, multiple time series can be filtered to identify a subset of time series that have time durations that exceed a preset time duration. Whether a time series of the subset of time series includes a time period with inactivity can be determined. Whether the time series exhibits a repetitive characteristic can be determined based on whether the time series has a pattern that repeats over a predetermined time period. Whether the time series includes a magnitude spike with a value above a preset magnitude can be determined. If the time series (i) lacks the time period with inactivity, (ii) exhibits the repetitive characteristic, and (iii) has the magnitude spike with the value above the preset magnitude threshold, the time series can be included in a data set for use with the three-stage predictor.

Abstract: In one embodiment, the present invention provides a neural network circuit comprising multiple symmetric core circuits. Each symmetric core circuit comprises a first core module and a second core module. Each core module comprises a plurality of electronic neurons, a plurality of electronic axons, and an interconnection network comprising multiple electronic synapses interconnecting the axons to the neurons. Each synapse interconnects an axon to a neuron. The first core module and the second core module are logically overlayed on one another such that neurons in the first core module are proximal to axons in the second core module, and axons in the first core module are proximal to neurons in the second core module. Each neuron in each core module receives axonal firing events via interconnected axons and generates a neuronal firing event according to a neuronal activation function.

Abstract: Disclosed is a method for the computer-assisted modeling of a technical system. One or more output vectors are modeled dependent on one or more input vectors by the learning process of a neural network on the basis of training data of known input vectors and output vectors. Each output vector comprises one or more operating variables of the technical system, and each input vector comprises one or more input variables that influence the operating variable(s). The neural network is a feedforward network with an input layer, a plurality of hidden layers, and an output layer. The output layer comprises a plurality of output clusters, each of which consists of one or more output neurons, the plurality of output clusters corresponding to the plurality of hidden layers. Each output cluster describes the same output vector and is connected to another hidden layer.

Abstract: A method, article comprising machine-readable instructions and apparatus that processes data systems for encoding, decoding, pattern recognition/matching and data generation is disclosed. State subsets of a data system are identified for the efficient processing of data based, at least in part, on the data system's systemic characteristics.

Abstract: A computer implemented method, a computerized system and a computer program product for generating questions. The computer implemented method comprising obtaining a question, wherein the question comprises one or more elements that define an answer for the question. The method further comprising obtaining the answer. The method further comprises automatically generating, by a processor, a new question based on the question and the answer. The automatic generation comprises determining a variant of the one or more elements, wherein the variant defines the answer, wherein the new question comprises the variant.

Abstract: Systems and methods are described for updating an influence model used to manage physical conditions of an environmentally controlled space. A method comprises operating an environmental maintenance system in a first production mode with the influence model until an event causes the system to enter a second production mode. In the second production mode a first actuator's operation level is varied and operation levels of other actuators are optimized. The influence model is adjusted based on the operation levels.

Abstract: A mechanism is provided for automatic genre determination of web content. For each type of web content genre, a set of relevant feature types are extracted from collected training material, where genre features and non-genre features are represented by tokens and an integer counts represents a frequency of appearance of the token in both a first type of training material and a second type of training material. In a classification process, fixed length tokens are extracted for relevant features types from different text and structural elements of web content. For each relevant feature type, a corresponding feature probability is calculated. The feature probabilities are combined to an overall genre probability that the web content belongs to a specific trained web content genre. A genre classification result is then output comprising at least one specific trained web content genre to which the web content belongs together with a corresponding genre probability.

Abstract: A circuit for implementing an artificial neuron comprises: an integrator for an input signal to produce a voltage signal; a signal generator linked to the integrator output producing two output signals when the voltage is at or above a predetermined voltage, a first signal for an output pulse of the neuron and a second signal for a control pulse; a resistive memory comprising two terminals switching from a high to low resistance state in a time following a statistical distribution specific to the memory, a first terminal linked to the output of the integrator; a transistor linked to a branch at zero potential to a second terminal of the resistive memory, controlled by the second output signal such that in the presence of a pulse of voltage the resistive memory switches from its high resistance state to its low resistance state with a view to lowering the voltage.

Abstract: The invention discloses the technology of brain-like computing virtualization. Brain-like computing means the computing technology to mimic human brain and generate human intelligence automatically with computer software. Here the unconscious engine and conscious engine are used to define human left and right brain, while the virtualization technology is used for software to run on future hardware, such as quantum computer and molecular computer. The applied domain areas include quantum gate and adiabatic quantum simulation, brain-like autonomic computing, traditional multi-core-cluster performance service, software development/service delivery systems, and mission-critical business continuity/disaster recovery.

Abstract: In one embodiment, prior to similarity measure computation, concept expansion is applied to an original ontology to generate an expanded ontology having the original concepts plus one or more pseudo-concepts, wherein at least one original concept is defined using a hierarchy of (possibly transitive) properties. As a result, the similarity measure computation can produce results that are better than those produced using conventional techniques. In one implementation, the similarity measure computation involves combining two similarity results: a first similarity result corresponding to common semantics found in the two concepts and a second similarity result corresponding to dissimilar semantics found in the two concepts.

Abstract: Information related to a time series can be predicted. For example, a repetitive characteristic of the time series can be determined by analyzing the time series for a pattern that repeats over a predetermined time period. An adjusted time series can be generated by removing the repetitive characteristic from the time series. An effect of a moving event on the adjusted time series can be determined. The moving event can occur on different dates for two or more consecutive years. A residual time series can be generated by removing the effect of the moving event from the adjusted time series. A base forecast that is independent of the repetitive characteristic and the effect of the moving event can be generated using the residual time series. A predictive forecast can be generated by including the repetitive characteristic and the effect of the moving event into the base forecast.

Abstract: A method, article comprising machine-readable instructions and apparatus that processes data systems for encoding, decoding, pattern recognition/matching and data generation is disclosed. State subsets of a data system are identified for the efficient processing of data based, at least in part, on the data system's systemic characteristics.

Abstract: Methods and systems for replacing feature values of features in training data with integer values selected based on a ranking of the feature values. The methods and systems are suitable for preprocessing large-scale machine learning training data.