Abstract: Systems and methods for predicting who is watching a program are disclosed. Text related to the program can be reviewed, the text comprising: plot information, sub-title information, summary information, script information, or synopsis information, or any combination thereof. Pre-determined genre words and pre-determined keywords can be determined based on machine learning analysis of historical programs. Words from the text which are relevant words can be determined, the relevant words being words that help identify genre words or keywords. How closely the relevant words coincide to the pre-determined genre words can be determined by generating a breakdown of how many relevant words are the pre-determined genre words and the pre-determined keywords. It can be predicted who will watch the program based on the breakdown.

Abstract: A method and system for determining predictably feasible model designs. The method includes defining a plurality of model designs, wherein the plurality of model designs include a plurality of infeasible model designs, wherein one or more of the infeasible model designs are infeasible due to limits in technology; storing information representing a plurality of technological trends; and classifying one or more of the infeasible model designs as predictably feasible model designs, wherein the predictable feasible model designs are those infeasible model designs expected to become feasible model designs if one or more of the plurality of technological trends continues as anticipated.

Abstract: Methods and apparatus are provided for memorizing data signals in a spiking neural network. For each data signal, such a method includes supplying metadata relating to the data signal to a machine learning model trained to generate an output signal, indicating a relevance class for a data signal, from input metadata for that data signal. The method includes iteratively supplying the data signal to a sub-assembly of neurons, interconnected via synaptic weights, of a spiking neural network and training the synaptic weights to memorize the data signal in the sub-assembly. The method further comprises assigning neurons of the network to the sub-assembly in dependence on the output signal of the model such that more relevant data signals are memorized by larger sub-assemblies. The data signal memorized by a sub-assembly can be subsequently recalled by activating neurons of that sub-assembly.

Abstract: A method, apparatus, and server for generating a user score based on social networking information is provided. In the disclosed method, by processing circuitry of an information processing apparatus, default annotation information of a plurality of sampled users, an ith user score and an ith relative user score for each of the sampled users are obtained. A user score model is trained according to the ith user score of the respective sampled user, the ith relative user score of the respective sampled user, and the default annotation information of the respective sampled user. An (i+1)th user score of the respective sampled user is subsequently calculated and a trained user score model, for each of the sampled users, is obtained when the (i+1)th user score for the respective sampled user satisfies a training termination condition, The method provides a solution to evaluate the user score for a use when personal information of the user is missing or incorrect.

Abstract: An exemplary embodiment may provide an interpretable neural network with hierarchical conditions and partitions. A local function f(x) may model the feature attribution within a specific partition. The combination of all the local functions creates a globally interpretable model. Further, INNs may utilize an external process to identify suitable partitions during their initialization and may support training using back-propagation and related techniques.

Abstract: A machine learning algorithm, such as a random forest regressor, can be trained using a set of annotated data objects to estimate the risk or business value for an object. The feature contributions for each data object can be analyzed and a representation generated that clusters data objects by feature contributions. Any clustering of data objects with incorrect scores in the visualization can be indicative of gaps in the regressor training. Adjustments to the inputs can be made, and the regressor retrained, to eliminate clustering of errors for similar feature contributions. Correcting the risk score estimations can ensure that the appropriate security policies and permissions are applied to each data object.

Abstract: A product configuration device outputs a configuration of a product in accordance with a set of configuration rules. The product configuration device includes a rule learning system configured to acquire a first set of data representing a plurality of configurations of the product; to generate a neural network model representing the first set of data; to extract relationships between configuration attributes from the neural network model; and to modify the set of configuration rules based on the extracted relationships to generate a modified set of configuration rules for the product configuration device. The product configuration device may also include a rule execution engine that outputs the configuration of the product based on the modified set of configuration rules.

Abstract: A computer-implemented method for managing datasets of a storage system is provided, wherein the datasets have respective sets of metadata, the method including: successively feeding first sets of metadata to a spiking neural network (SNN), the first sets of metadata fed corresponding to datasets of the storage system that are labeled with respect to classes they belong to, so as to be associated with class labels, for the SNN to learn representations of said classes in terms of connection weights that weight the metadata fed; successively feeding second sets of metadata to the SNN, the second sets of metadata corresponding to unlabeled datasets of the storage system, for the SNN to infer class labels for the unlabeled datasets, based on the second sets of metadata fed and the representations learned; and managing datasets in the storage system, based on class labels of the datasets, these including the inferred class labels.

Abstract: Techniques for classifying heartbeats using patient electrocardiogram (ECG) data are described. ECG data is received, including waveform data and time interval data relating to a plurality of heartbeats for the patient. A convolutional neural network in a first path of a machine learning architecture generates a first plurality of output values by analyzing the waveform data. A fully-connected neural network in a second path of the machine learning architecture generates a second plurality of output values by analyzing the time interval data. The plurality of heartbeats in the ECG data are classified by concatenating the first plurality of output values and the second plurality of output values using the machine learning architecture.

Abstract: A machine receives a first set of global parameters from a global parameter server. The first set of global parameters includes data that weights one or more operands used in an algorithm that models an entity type. Multiple learner processors in the machine execute the algorithm using the first set of global parameters and a mini-batch of data known to describe the entity type. The machine generates a consolidated set of gradients that describes a direction for the first set of global parameters in order to improve an accuracy of the algorithm in modeling the entity type when using the first set of global parameters and the mini-batch of data. The machine transmits the consolidated set of gradients to the global parameter server. The machine then receives a second set of global parameters from the global parameter server, where the second set of global parameters is a modification of the first set of global parameters based on the consolidated set of gradients.

Abstract: A computation method used in a convolutional neural network is provided. The method includes: receiving original data; determining a first optimal quantization step size according to a distribution of the original data; performing fixed-point processing to the original data according to the first optimal quantization step size to generate first data; inputting the first data to a first layer of the convolutional neural network to generate first output data; determining a second optimal quantization step size according to a distribution of the first output data; performing the fixed-point processing to the first output data according to the second optimal quantization step size to generate second data; and inputting the second data to a second layer of the convolutional neural network.

Abstract: Methods, systems and computer program products implementing character-level deep neural networks for information extraction are disclosed. A system uses character-level information retrieved from a transaction record to classify the transaction as a whole and to tag individual sections of the transaction record by entity type. The system processes the transaction record using multiple and separate character-level models. The system can use a one-dimensional neural network for featurization fed into a fully connected network for classification for identifying the most common classes of a transaction record. The system can identify one or more entities, e.g., service provider names, from the transaction using an RNN. The RNN can include one or more LSTM models. The LSTM models can be BI-LSTM models.

Abstract: Ancestry deconvolution includes obtaining unphased genotype data of an individual; phasing, using one or more processors, the unphased genotype data to generate phased haplotype data; using a learning machine to classify portions of the phased haplotype data as corresponding to specific ancestries respectively and generate initial classification results; and correcting errors in the initial classification results to generate modified classification results.

Abstract: An approach for continuously provisioning machine learning models, executed by one or more computer nodes to provide a future prediction in response to a request from one or more client devices, is provided. The approach generates, by the one or more computer nodes, a machine learning model. The approach determines, by the one or more computer nodes, whether the machine learning model is a new model. In response to determining the machine learning model is not the new model, the approach retrieves, by the one or more computer nodes, one or more model containers with an associated model to a new persistent model. The approach determines, by the one or more computer nodes, a difference between the associated model and the new persistent model. Further, in response to determining the machine learning model is the new model, the approach generates, by the one or more computer nodes, one or more model containers.

Abstract: An online system receives explicit user data and explicit event data, and implicit user data and implicit event data from a third party system. The online system generates an implicit users/implicit events data feature, an explicit users/explicit events data feature, and an explicit users/implicit events data feature. The online system generates a prediction of the counterfactual rate based on the implicit users/implicit events data feature, the explicit users/explicit events data feature, and the explicit users/explicit events data feature, the counterfactual rate indicating the likelihood that target users matching certain characteristics caused an event to occur when the target are not been presented with content by the online system, the content configured to induce users to cause the event to occur. A combined prediction rate is presented to the third party system based on the counterfactual rate.

Abstract: A method and apparatus that detect wheel misalignment are provided. The method includes predicting a self-aligning torque parameter based on a regression model determined from a dataset including one or more from among a steering wheel angle parameter, a speed parameter, a torsion bar torque parameter, a lateral acceleration parameter, and a power steering torque parameter, comparing a measured self-aligning torque parameter and the predicted self-aligning torque parameter, and outputting a wheel alignment condition indicating whether the wheel alignment is proper if the self-aligning torque parameter and the predicted self-aligning torque parameter are within a predetermined value based on the comparing.

Abstract: A processing unit can extract salient semantics to model knowledge carryover, from one turn to the next, in multi-turn conversations. Architecture described herein can use the end-to-end memory networks to encode inputs, e.g., utterances, with intents and slots, which can be stored as embeddings in memory, and in decoding the architecture can exploit latent contextual information from memory, e.g., demographic context, visual context, semantic context, etc. e.g., via an attention model, to leverage previously stored semantics for semantic parsing, e.g., for joint intent prediction and slot tagging. In examples, architecture is configured to build an end-to-end memory network model for contextual, e.g., multi-turn, language understanding, to apply the end-to-end memory network model to multiple turns of conversational input; and to fill slots for output of contextual, e.g., multi-turn, language understanding of the conversational input.

Abstract: Some embodiments of the invention provide a novel method for training a multi-layer node network that mitigates against overfitting the adjustable parameters of the network for a particular problem. During training, the method of some embodiments adjusts the modifiable parameters of the network by iteratively identifying different interior-node, influence-attenuating masks that effectively specify different sampled networks of the multi-layer node network. An interior-node, influence-attenuating mask specifies attenuation parameters that are applied (1) to the outputs of the interior nodes of the network in some embodiments, (2) to the inputs of the interior nodes of the network in other embodiments, or (3) to the outputs and inputs of the interior nodes in still other embodiments. In each mask, the attenuation parameters can be any one of several values (e.g., three or more values) within a range of values (e.g., between 0 and 1).

Abstract: Operations of computing devices are managed using one or more deep neural networks (DNNs), which may receive, as DNN inputs, data from sensors, instructions executed by processors, and/or outputs of other DNNs. One or more DNNs, which may be generative, can be applied to the DNN inputs to generate DNN outputs based on relationships between DNN inputs. The DNNs may include DNN parameters learned using one or more computing workloads. The DNN outputs may be, for example, control signals for managing operations of computing devices, predictions for use in generating control signals, warnings indicating an acceptable state is predicted, and/or inputs to one or more neural networks. The signals enhance performance, efficiency, and/or security of one or more of the computing devices. DNNs can be dynamically trained to personalize operations by updating DNN weights or other parameters.

Abstract: Apparatuses and methods of manufacturing same, systems, and methods for performing network parameter quantization in deep neural networks are described. In one aspect, diagonals of a second-order partial derivative matrix (a Hessian matrix) of a loss function of network parameters of a neural network are determined and then used to weight (Hessian-weighting) the network parameters as part of quantizing the network parameters. In another aspect, the neural network is trained using first and second moment estimates of gradients of the network parameters and then the second moment estimates are used to weight the network parameters as part of quantizing the network parameters. In yet another aspect, network parameter quantization is performed by using an entropy-constrained scalar quantization (ECSQ) iterative algorithm. In yet another aspect, network parameter quantization is performed by quantizing the network parameters of all layers of a deep neural network together at once.