Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for using a quantum oracle to make inference in complex machine learning models that is capable of solving artificial intelligent problems. Input to the quantum oracle is derived from the training data and the model parameters, which maps at least part of the interactions of interconnected units of the model to the interactions of qubits in the quantum oracle. The output of the quantum oracle is used to determine values used to compute loss function values or loss function gradient values or both during a training process.

Abstract: Systems and methods for reducing the size of neural networks are disclosed. In an embodiment, a server computer stores a plurality of training datasets, each of which comprise a plurality of training input matrices and a plurality of corresponding outputs. The server computer initiates training of a neural network using the plurality of training input matrices, a weight matrix, and the plurality of corresponding outputs. While the training of the neural network is being performed, the server computer identifies one or more weight values of the weight matrix for removal. The server computer removes the one or more weight values from the weight matrix to generate a reduced weight matrix. The server computer then stores the reduced weight matrix with the neural network.

Abstract: There is provided a classifier capable of classifying unknown abnormal data input to the classifier even if there is a small number of abnormal data used for the learning of the classifier. When learning parameters of the classifier, the specific category likelihood of normal patterns for learning relatively deviating from a group of normal patterns for learning is decreased relatively to the specific category likelihood of normal patterns for learning not relatively deviating from the group of normal patterns for learning, and the specific category likelihood of abnormal patterns for learning is decreased relatively to the specific category likelihood of the group of normal patterns for learning.

Abstract: Data is classified using semi-supervised data. A decomposition is performed to define a first decomposition matrix that includes first eigenvectors of a weight matrix, a second decomposition matrix that includes second eigenvectors of a transpose of the weight matrix, and a diagonal matrix that includes eigenvalues of the first eigenvectors. Eigenvectors are selected from the first eigenvectors to define a reduced decomposition matrix. A linear transformation matrix is computed as a function of the first decomposition matrix, the reduced decomposition matrix, the diagonal matrix, and a penalty matrix. When a rank of the linear transformation matrix is less than a number of rows of the penalty matrix, a classification matrix is computed by updating a gradient of a cost function. When the rank of the linear transformation matrix is equal to the number of rows of the penalty matrix, the classification matrix is computed using a dual formulation.

Abstract: Dynamic gating for neuromorphic systems and the configuration thereof are provided. In various embodiments, neurosynaptic system comprises a neurosynaptic core. The neuromorphic core comprises a plurality of neurons and axons. The neurosynaptic core comprises a programmable gate operative to receive a control signal and selectively output a first output signal based on the control signal. In various embodiments, a plurality of input parameters are read, defining the behavior of a programmable gate. Based upon the plurality of input parameters, a neurosynaptic core is configured to provide a programmable gate operative to receive a control signal and selectively output a first output signal based on the control signal.

Abstract: The present disclosure is directed to reducing model size of a machine learning model with encoding. The input to a machine learning model may be encoded using a probabilistic data structure with a plurality of mapping functions into a lower dimensional space. Encoding the input to the machine learning model results in a compact machine learning model with a reduced model size. The compact machine learning model can output an encoded representation of a higher-dimensional space. Use of such a machine learning model can include decoding the output of the machine learning model into the higher dimensional space of the non-encoded input.

Abstract: Provided is an apparatus including a processor and one or more computer readable mediums collectively including instructions that, when executed by the processor, cause the processor to: acquire a candidate for a solution of an optimization problem for optimizing a third objective function based on a first objective function and a second objective function; obtain, as another candidate for the solution of the optimization problem, a solution that optimizes the second objective function under a constraint corresponding to a value of the first objective function for the acquired candidate; and select the solution of the optimization problem from among the plurality of candidates for the solution of the optimization problem. Also provided as the first aspect are a method and non-transitory computer readable storage medium.

Abstract: Methods, systems, and apparatus for solving optimization tasks. In one aspect, a method includes receiving input data comprising (i) data specifying an optimization task to be solved, and (ii) data specifying task objectives for solving the optimization task, comprising one or more local task objectives and one or more global task objectives; processing the received input data to obtain one or more initial solutions to the optimization task based on the local task objectives, wherein at least one initial solution is obtained from a first quantum computing resource; and processing the generated one or more initial solutions using a second quantum computing resource to generate a global solution to the optimization task based on the global task objectives.

Abstract: Disclosed herein is a system and method that can be used with any underlying classification technique. The method receives a test dataset and determines the features in that test dataset that are present. From these features the training dataset is modified to only have those features that are present in the test dataset. This modified test dataset is then used to calibrate the classifier for the particular incoming data set. The process repeats itself for each different incoming dataset providing a just in time calibration of the classifier.

Abstract: Embodiments of the present invention may provide the capability to detect complex events while providing improved detection and performance. In an embodiment of the present invention, a method for detecting an event may comprise receiving data representing measurement or detection of physical parameters, conditions, or actions, quantizing the received data and selecting a number of samples from the quantized data, generating a hidden Markov model representing events to be detected using initial model values based on ideal conditions, wherein a desired output is defined as a sequence of states, and wherein a number of states of the hidden Markov model is less than or equal to the number of samples of the quantized data, adjusting the quantized data and the initial model values to improve accuracy of the model, determining a state sequence of the hidden Markov model, and outputting an indication of a detected event.

Abstract: Systems and methods for anomaly detection in complex physical systems, including extracting features representative of a temporal evolution of the complex physical system, and analyzing the extracted features by deriving vector trajectories using sliding window segmentation of time series, applying a linear test to determine whether the vector trajectories are linear, and performing subspace decomposition on the vector trajectory based on the linear test. A system evolution model is generated from an ensemble of models, and a fitness score is determined by analyzing different data properties of the system based on specific data dependency relationships. An alarm is generated if the fitness score exceeds a predetermined number of threshold violations for the different data properties.

Abstract: An information delivery system allows for the organization and presentation of information to users. Illustratively, aspects of the disclosure correspond to a system and method which provides for interactive information delivery, or interactive learning. More particularly, a platform is disclosed which provides an independent interactive interface for content delivery and e-learning and for creation of teaching or learning presentations.

Abstract: A neural network processing system includes at least one synapse and a neuron circuit. The synapse receives an input signal and has an external weighted value and an internal weighted value, and the internal weighted value has a variation caused by an external stimulus. When the variation of the internal weighted value accumulates to a threshold value, the external weighted value varies and the input signal is multiplied by the external weighted value of the synapse to generate a weighted signal. A neuron circuit is connected with the synapse to receive the weighted signal transmitted by the synapse, and calculates and outputs the weighted signal. The present invention can simultaneously accelerate the prediction and learning functions of the deep learning and realize a hardware neural network with high precision and real-time learning.

Abstract: Described is neuromorphic system for authorized user detection. The system includes a client device comprising a plurality of sensor types providing streaming sensor data and one or more processors. The one or more processors include an input processing component and an output processing component. A neuromorphic electronic component is embedded in or on the client device for continuously monitoring the streaming sensor data and generating out-spikes based on the streaming sensor data. Further, the output processing component classifies the streaming sensor data based on the out-spikes to detect an anomalous signal and classify the anomalous signal.

Abstract: The automatic selection and usage of a parser is disclosed. Raw data is obtained from a first remote device. At least a portion of the raw data is evaluated using a plurality of rules. A confidence measure is determined for at least some of the rules. An indication that the raw data pertains to a source is provided as output when the confidence measure exceeds a threshold.

Abstract: A system to reduce the number of factors that need to be considered in generating a prediction function includes an access module and a function generator module. The access module accesses a reduced set of factors derived from an original set of factors based at least in part on correlations between the factors of the original set. The function generator module generates, based on the reduced set of factors and a data set associated therewith, a plurality of potential prediction functions that operate on the data set to predict a result, evaluates performance of each one from the plurality of potential prediction functions, and selects a solution prediction function based on the evaluated.

Abstract: A networked system for managing a physical intrusion detection/alarm includes tiers devices and a rules engine and router to interact with the rules engine and rule engine results, where the router is configured to feed inputs to and receive outputs from the rules engine, and the router further configured to programmatically route results of rule execution by the rules engine to a hierarchical structure stored in computer storage for access by subscriber devices.

Abstract: Embodiments relate to a system, program product, and method for use with an intelligent computer platform to decipher analogical phrases. A phrase is parsed into a set of terms to reveal an analogical pattern. The set of terms are categorized according to syntactic placement and each term is placed into two or more categories according to word type patterns in the phrase to produce metadata. The metadata is matched to outcome metadata generated from a set of outcomes produced from data storage and a set of grammatical data is generated for each potential outcome. A statistical model is trained, whereby the training includes weighing and ranking each potential outcome according to degree of congruence with the syntactic placement and word type patterns of the phrase. A highest outcome is selected, the highest outcome being the potential outcome with the highest rank and a confidence level data metric is then applied.

Abstract: A gesture classification apparatus and method is disclosed. The apparatus may include a feature extractor configured to extract a plurality of features using a electromyogram (EMG) data group obtained from an EMG signal sensor including a plurality of channels, an artificial neural network including an input layer to which the EMG data group corresponding to the plurality of features is input and an output layer configured to output a preset gesture corresponding to the plurality of features, and a gesture recognizer configured to recognize a gesture performed by a user and corresponding to the extracted features.

Abstract: Embodiments of this invention comprise modeling a subject's state and the influence of training treatments, or actions, on that state to create a training policy. Both state and effects of actions are modeled as probabilistic using Partially Observable Markov Decision Process (POMDP) techniques. Utilizing this model and the resulting training policy with subjects creates an effective decision aid for instructors to improve learning relative to a traditional scenario selection strategy.