Abstract: At a network-accessible artificial intelligence service for generating content-based recommendations based on multi-perspective learned descriptors, text sections associated with a plurality of description perspectives, including a single-character perspective and a multi-character perspective, are extracted from various text sources. Using the text sections as input, a machine learning model which includes respective portions corresponding to the different perspectives is trained to reconstruct the input using intermediary descriptors learned from the input. An indication that a second text source is recommended with respect to a first text source is generated using a set of the learned descriptors and transmitted.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for performing a machine learning task on a network input to generate a network output. In one aspect, one of the systems includes an attention neural network configured to perform the machine learning task, the attention neural network including one or more LSH attention layers, each LSH attention layer comprising one or more LSH attention sub-layers, each LSH sub-layer configured to: receive a sequence of queries derived from an input sequence to the LSH attention layer, the sequence of queries having a respective query at each of a plurality of input positions; determine one or more respective hash values for each of the respective queries at each of the plurality of input positions; generate a plurality of LSH groupings; and generate an attended input sequence.

Abstract: Methods and systems for generating a multi-model ensemble of global climate simulation data from a plurality of global climate simulation models (GCMs), to be used in training a neural network (NN)-based climate forecasting model, are disclosed. The methods and systems perform steps of computing a GCM validation measure for each GCM; selecting a validated subset of the GCMs, by comparing each computed GCM validation measure to a validation threshold determined based on observational historical climate data; computing a forecast skill score for each validated GCM, based on a first forecast function; selecting a validated and skillful subset of GCMs; generating one or more candidate ensembles by combining simulation data from at least two validated and skillful GCMs; computing an ensemble forecast skill score for each candidate ensemble, based on a second forecast function; and selecting a best-scored candidate ensemble.

Abstract: A commercial process with a dependent variable can be associated with a set of independent variables. The commercial process can continuously provide data collection opportunities. An intervention is designed using a model to predict the dependent outcome. The actual outcome of the intervention can be determined within the window of utility for these data. One objective is to improve intervention outcomes with prediction. Purely random outcomes (no model prediction) and outcomes resulting from the intervention (model operations) are aggregated into separate files—a sequence of control model data files and a sequence of model data files of operational data. These model data files and control model data files are used to analyze model performance and to react automatically when identified conditions warrant.

Abstract: A method and an apparatus for recommending a message. The method for recommending a message in the present disclosure includes separately parsing a first message published by a first user on a network and a second message published by a second user on the network, obtaining interest description information of the first message and topic description information of the second message, where the second user is another user except the first user, comparing the topic description information with the interest description information, and calculating a similarity of the topic description information and the interest description information; and if the similarity is greater than or equal to a predetermined value, pushing the second message published by the second user to the first user. A user can conveniently and flexibly obtain a message in which the user is interested in the embodiments of the present disclosure.

Abstract: A method, system and computer readable medium for generating a cognitive insight comprising: receiving data, the data comprising a plurality of examples, each of the plurality of examples comprising an input object and a desired output value, at least some of the plurality of examples being based upon feedback from a user; performing a machine learning operation on the data, the machine learning operation comprising performing an augmented gamma belief network operation, the augmented gamma belief network operation producing an inferred function based upon the data; and, generating a cognitive insight based upon the cognitive profile generated using the inferred function generated by the augmented gamma belief network operation.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for augmenting neural networks with an external memory. One of the systems includes a controller neural network that includes a Least Recently Used Access (LRUA) subsystem configured to: maintain a respective usage weight for each of a plurality of locations in the external memory, and for each of the plurality of time steps: generate a respective reading weight for each location using a read key, read data from the locations in accordance with the reading weights, generate a respective writing weight for each of the locations from a respective reading weight from a preceding time step and the respective usage weight for the location, write a write vector to the locations in accordance with the writing weights, and update the respective usage weight from the respective reading weight and the respective writing weight.

Abstract: In an approach, a computer determines based, at least in part, on a video of an attendee of a video conference, a first sentiment of the attendee wherein the first sentiment includes at least a sentiment from a sentiment analysis of one or more facial expressions of the attendee and a sentiment from a sentiment analysis of a plurality of the attendee's spoken words. The approach includes a computer receiving an indication of an attendee activity in at least a first application in computing devices accessed by the attendee and determining whether the first sentiment of the attendee is related to the video conference based, in part, on the attendee activity in at least the first application. Responsive to determining that the first sentiment of the attendee is not related to the video conference, the computer discards the first sentiment that is unrelated to the video conference.

Abstract: In certain embodiments, training data may be generated for training a prediction model. Training data including first datasets may be obtained, where the first datasets include a plurality of feature types. A determination, via a relevancy model, based on the training data, of whether a feature type satisfies a first condition may be made. If the first condition is satisfied, one or more second datasets may be obtained to update the training data, where the second datasets include the plurality of feature types. A determination, via the relevancy model, based on the updated training data, may be made as to whether the feature type satisfies a second condition. The first and second conditions may relate to whether the feature type has a threshold amount of influence on the prediction model. If the second condition is satisfied, the updated training data may be provided to the prediction model.

Abstract: A method for adapting the operation of an apparatus connected to a network deployed in an environment, including the steps of: collecting data relating to the environment from a plurality of sources; identifying usual environmental situations from an analysis of the collected data; detecting a deviation from at least one identified usual situation; and sending a control to the apparatus for adapting the operation thereof to the detected deviation.

Abstract: A system including one or more computers and one or more storage devices storing instructions that when executed by the one or more computers cause the one or more computers to implement a memory and memory-based neural network is described. The memory is configured to store a respective memory vector at each of a plurality of memory locations in the memory. The memory-based neural network is configured to: at each of a plurality of time steps: receive an input; determine an update to the memory, wherein determining the update comprising applying an attention mechanism over the memory vectors in the memory and the received input; update the memory using the determined update to the memory; and generate an output for the current time step using the updated memory.

Abstract: According to an embodiment, a training device trains a neural network that outputs a posterior probability that an input signal belongs to a particular class. An output layer of the neural network includes N units respectively corresponding to classes and one additional unit. The training device includes a propagator, a probability calculator, and an updater. The propagator supplies a sample signal to the neural network and acquires (N+1) input values for each unit at the output layer. The probability calculator supplies the input values to a function to generate a probability vector including (N+1) probability values respectively corresponding to the units at the output layer. The updater updates a parameter included in the neural network in such a manner to reduce an error between a teacher vector including (N+1) target values and the probability vector. A target value corresponding to the additional unit is a predetermined constant value.

Abstract: An illustrative embodiment of a computer-implemented process for non-leading computer aided detection of features of interest in a dataset, designates a particular formation using a computer recognizable gesture to identify a gestured location in an analyzed view of the dataset in response to a user identifying the particular formation in the analyzed view. The dataset is generated by a computer and representative of a portion of an object characterized by the dataset. Responsive to identifying the gestured location, the particular formation is displayed to the user, and a composition is revealed including additional structural imagery, functional imagery and findings resulting from machine learning and analysis. Responsive to revealing the composition to the user, the user is prompted to select performance of accept selection, reject selection or modify selection with regard to the particular formation displayed.

Abstract: Methods and arrangements for managing data collection relating to equipment in an asset network. A model is trained based on historical data relative to equipment in the asset network, wherein the model is employed for recommending at least one action for further data collection from the equipment. The model is adapted based on inputs comprising two or more of: historical sensor data, from one or more sensors obtaining data relative to the equipment; equipment maintenance and/or replacement data; a current system state; and a geographical position of one or more individuals who report data. The adapted model is employed to recommend at least one subsequent best action for collecting data relative to the equipment. Other variants and embodiments are broadly contemplated herein.

Abstract: The technology disclosed provides a so-called “joint many-task neural network model” to solve a variety of increasingly complex natural language processing (NLP) tasks using growing depth of layers in a single end-to-end model. The model is successively trained by considering linguistic hierarchies, directly connecting word representations to all model layers, explicitly using predictions in lower tasks, and applying a so-called “successive regularization” technique to prevent catastrophic forgetting. Three examples of lower level model layers are part-of-speech (POS) tagging layer, chunking layer, and dependency parsing layer. Two examples of higher level model layers are semantic relatedness layer and textual entailment layer. The model achieves the state-of-the-art results on chunking, dependency parsing, semantic relatedness and textual entailment.

Abstract: Disclosed is a novel system, computer program product, and method to compute an air quality forecast. An air quality forecast model, air quality real-time monitoring data, and air quality forecast data is accessed. A deviation in air pollution emission is monitored by classifying a difference between the air quality monitoring data and the air quality forecast data. This monitoring includes classifying any weather differences which are caused by weather, classifying any terrain differences which are caused by a geographic terrain; and, filtering the difference caused by inaccurate pollution emission inventory. The monitoring may repeat in response to a given time period elapsing or a chance in air quality forecast data received.

Abstract: Methods and apparatus for extending a neural network, reducing its dimension and processing input data are provided. The method of extending a neural network involves selecting, with a processor, a node of a neural network, adding a new node in a layer that includes the selected node, and setting connection weights of the new node based on connection weights of the selected node.

Abstract: Mechanisms are provided to implement a medical condition base cartridge generator. The mechanisms ingest an electronic corpus of medical content and generate a medical condition base cartridge for a medical condition. The medical condition base cartridge is a pluggable cartridge comprising insight data structures that specify an association of clinical attributes of patients with the medical condition and a treatment for the medical condition. The mechanisms install the medical condition base cartridge as a resource for performing a cognitive operation.

Abstract: Methods, systems, and computer program products for expressive temporal predictions over semantically-driven time windows are provided herein. A computer-implemented method includes identifying, within a knowledge graph pertaining to a given prediction, a subset of the knowledge graph related to one or more predicted training examples, wherein the subset comprises (i) a set of nodes and (ii) one or more relationships among the set of nodes; determining, for the identified subset, one or more snapshots of the knowledge graph relevant to the given prediction; quantifying a validity window for the one or more predicted training examples, wherein the validity window comprises a temporal bound for prediction validity; and computing a validity window for the given prediction based on the quantified validity window for the one or more predicted training examples.

Abstract: Method embodiments and/or system embodiments are provided that may be utilized to recommend online content to users based, at least in part on a prediction of diffusion of online content through a social network.