Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for implementing long-short term memory layers with compressed gating functions. One of the systems includes a first long short-term memory (LSTM) layer, wherein the first LSTM layer is configured to, for each of the plurality of time steps, generate a new layer state and a new layer output by applying a plurality of gates to a current layer input, a current layer state, and a current layer output, each of the plurality of gates being configured to, for each of the plurality of time steps, generate a respective intermediate gate output vector by multiplying a gate input vector and a gate parameter matrix. The gate parameter matrix for at least one of the plurality of gates is a structured matrix or is defined by a compressed parameter matrix and a projection matrix.

Abstract: Systems and methods are disclosed for selecting cohorts. In one implementation, a model-assisted selection system for identifying candidates for placement into a cohort includes a data interface and at least one processing device. The at least one processing device is programmed to access, via the data interface, a database from which feature vectors associated with an individual from among a population of individuals can be derived; derive, for the individual, one or more feature vectors from the database; provide the one or more feature vectors to a model; receive an output from the model; and determine whether the individual from among the population of individuals is a candidate for the cohort based on the output received from the model.

Abstract: A system, method and a computer program product may be provided for automatically creating and parameterizing a semantically-enriched diagnosis model for an entity. The system receives a list of data points, from sensors or a database, to be used to create a diagnosis model. The system automatically creates the diagnosis model based on the received list of data points and data stored in a database and parameterizes the diagnosis model. The parameterized diagnosis model reflects rules that determine one or more potential causes of one or more abnormalities of one or more physical conditions in the entity.

Abstract: A question database storing questions and a conditional probability of one question to be asked given that a previous question was asked is searched to predict a future question based on the conditional probability stored in the question database given an input question as the previous question. The future question is suggested to a user. Responsive to receiving an acceptance of the future question, the question database is updated to strengthen the conditional probability associated with the future question occurring given the input question. An answer to the future question can be provided and searching, predicting, suggesting and updating may be repeated, with the future question as the input question, until the future question is declined.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for performing parallel generation of output from an autoregressive sequence to sequence model. In one aspect, a blockwise parallel decoding method takes advantage of the fact that some architectures can score sequences in sublinear time. By generating predictions for multiple time steps at once then backing off to a longest prefix validated by the scoring model, the methods can substantially improve the speed of greedy decoding without compromising performance.

Abstract: Technology related to training a neural network accelerator using mixed precision data formats is disclosed. In one example of the disclosed technology, a neural network accelerator is configured to accelerate a given layer of a multi-layer neural network. An input tensor for the given layer can be converted from a normal-precision floating-point format to a quantized-precision floating-point format. A tensor operation can be performed using the converted input tensor. A result of the tensor operation can be converted from the block floating-point format to the normal-precision floating-point format. The converted result can be used to generate an output tensor of the layer of the neural network, where the output tensor is in normal-precision floating-point format.

Abstract: Systems and methods for improvements in AI model learning and updating are provided. The model updating may reuse existing business conversations as the training data set. Features within the dataset may be defined and extracted. Models may be selected and parameters for the models defined. Within a distributed computing setting the parameters may be optimized, and the models deployed. The training data may be augmented over time to improve the models. Deep learning models may be employed to improve system accuracy, as can active learning techniques. The models developed and updated may be employed by a response system generally, or may function to enable specific types of AI systems. One such a system may be an AI assistant that is designed to take use cases and objectives, and execute tasks until the objectives are met. Another system capable of leveraging the models includes an automated question answering system utilizing approved answers.

Abstract: A method formats a weight matrix in a current layer included in a neural network. The method includes calculating a row length for each row of the weight matrix based on a number of elements each of which has non-zero value; storing rearrangement information including result of sorting rows in the order of row lengths; performing a row transformation or the row transformation and a column transformation on the weight matrix using the rearrangement information; distributing rows of a transformed weight matrix to a plurality of processing elements (PEs); and generating formatted data including one or more group data each including values and column information being processed in each of the PEs.

Abstract: A model calculating unit for calculating a neural layer of a multilayer perceptron model having a hardwired processor core developed in hardware for calculating a definitely specified computing algorithm in coupled functional blocks. The processor core is designed to calculate, as a function of one or multiple input variables of an input variable vector, of a weighting matrix having weighting factors and an offset value specified for each neuron, an output variable for each neuron for a neural layer of a multilayer perceptron model having a number of neurons, a sum of the values of the input variables weighted by the weighting factor, determined by the neuron and the input variable, and the offset value specified for the neuron being calculated for each neuron and the result being transformed using an activation function in order to obtain the output variable for the neuron.

Abstract: A search point determining method in an estimation process of a function, executed by a processor included in a search point determining apparatus, the method includes, calculating a search prediction time and a confidence interval upper limit obtained by using a Gaussian process for the function in each search candidate point from a past search result of the function, generating an area in a parameter space for each search candidate point by using a position of a search point close to the relevant search candidate point in a past search result, a search prediction time corresponding to each search candidate point, and a confidence interval upper limit corresponding to each search candidate point, and determining a search point based on a size of the area in a plurality of parameter spaces.

Abstract: The present disclosure provides a system for improving performance of a neural network model. The system receives the neural network model and a training data associated with the neural network model. In addition, the system examines a first plurality of neuron activations inside the neural network model for the training data. The system examines the first plurality of neurons for creating a statistical profile of the first plurality of neuron activations. Further, the system receives a new set of data samples to improve the neural network model. Furthermore, the system examines a second plurality of neuron activations of each new sample of the new set of data samples. Moreover, the system extracts one or more data samples from the new set of data samples with largest novelty measurements. Also, the system adds the extracted one or more samples to the training data for re-training of the neural network model.

Abstract: Approaches for multi-hop knowledge graph reasoning with reward shaping include a system and method of training a system to search relational paths in a knowledge graph. The method includes identifying, using an reasoning module, a plurality of first outgoing links from a current node in a knowledge graph, masking, using the reasoning module, one or more links from the plurality of first outgoing links to form a plurality of second outgoing links, rewarding the reasoning module with a reward of one when a node corresponding to an observed answer is reached, and rewarding the reasoning module with a reward identified by a reward shaping network when a node not corresponding to an observed answer is reached. In some embodiments, the reward shaping network is pre-trained.

Abstract: A device (DS) monitors a process using at least one electronic device (EE1-EE4) in operation and generating first data of a metric. This device (DS) comprises: learning means (MA) configured to analyse automatically second data which are representative of events that have occurred in the course of the process, in order to determine anomalies of a chosen type, and then automatically determine an indicator representative of this metric, then a correlation between these determined anomalies and this indicator, and then at least one rule defining this correlation, and monitoring means (MS1) configured to analyse newly generated first data periodically, and group by group, by checking whether at least one value of the indicator determined on the basis of the aforesaid data satisfies this determined rule, in order to predict the occurrence of the anomaly in a future group of first data when this at least one value satisfies this rule.

Abstract: A system and method for generating an interface for providing recommendations based on contextual insights, the method including: generating at least one signature for at least one multimedia content element identified within an interaction between a plurality of users; generating at least one contextual insight based on the generated at least one signature and user interests of the plurality of users, wherein each contextual insight indicates a current user preference; searching for at least one content item that matches the at least one contextual insight; and generating an interface for providing the at least one content item within the interaction between the plurality of users.

Abstract: A method for biomarker discovery for substance use disorders wherein high dimensional data containing a plurality of variables based on a sample size and a number of variables exceeding that sample size are applied to an ensemble of statistical learning models whereby biomarkers of a substance use disorder are identified.

Abstract: A system for bit-serial computation in a neural network is described. The system may be embodied on an integrated circuit and include one or more bit-serial tiles for performing bit-serial computations in which each bit-serial tile receives input neurons and synapses, and communicates output neurons. Also included is an activation memory for storing the neurons and a dispatcher and a reducer. The dispatcher reads neurons and synapses from memory and communicates either the neurons or the synapses bit-serially to the one or more bit-serial tiles. The other of the neurons or the synapses are communicated bit-parallelly to the one or more bit-serial tiles, or according to a further embodiment, may also be communicated bit-serially to the one or more bit-serial tiles. The reducer receives the output neurons from the one or more tiles, and communicates the output neurons to the activation memory.

Abstract: A processing unit implements a convolutional neural network (CNN) by fusing at least a portion of a convolution phase of the CNN with at least a portion of a batch normalization phase. The processing unit convolves two input matrices representing inputs and weights of a portion of the CNN to generate an output matrix. The processing unit performs the convolution via a series of multiplication operations, with each multiplication operation generating a corresponding submatrix (or “tile”) of the output matrix at an output register of the processing unit. While an output submatrix is stored at the output register, the processing unit performs a reduction phase and an update phase of the batch normalization phase for the CNN. The processing unit thus fuses at least a portion of the batch normalization phase of the CNN with a portion of the convolution.

Abstract: Disclosed herein are methods, systems, and processes to automate cluster interpretation in computing environments to develop targeted remediation security actions. To interpret clusters that are generated by a clustering methodology without subjecting clustered data to classifier-based processing, separation quantifiers that indicate a spread in feature values across clusters are determined and used to discover relative feature importances of features that drive the formation of clusters, permitting a security server to identify features that discriminate between clusters.

Abstract: The present technology addresses the problem of quickly and safely improving policies in online reinforcement learning domains. As its solution, an exploration strategy comprising diverse exploration (DE) is employed, which learns and deploys a diverse set of safe policies to explore the environment. DE theory explains why diversity in behavior policies enables effective exploration without sacrificing exploitation. An empirical study shows that an online policy improvement algorithm framework implementing the DE strategy can achieve both fast policy improvement and safe online performance.

Abstract: Embodiments of the present invention disclose a method, a computer program product, and a computer system predicting parking availability. A computer identifies parking spaces and groups the parking spacing into parking locations. In addition, the computer distinguishes private parking spaces from public parking spaces, and trains a crowd forecast model for each of the parking locations. The computer further receives a destination and preferences, from which the computer creates a geofence based on the destination and preferences. The computer then predicts parking availability based on the crowd forecast models and refines the crowd forecast model.