Abstract: A device for estimating a cause of an anomaly comprises: a detection unit to detect an anomaly in a detection target based on a learner trained on first numerical vectors obtained from a detection target when the detection target is under a normal condition and second numerical vectors to be obtained from the detection target at multiple time; and a first computing unit to compute, for each metric of a second numerical vector from which an anomaly has been detected, as information for estimating a metric of cause of the anomaly, a value obtained by subtracting, from a value of the metric, an average of the metric in the first numerical vectors, and dividing a result of the subtracting by standard deviation of the metric in the first numerical vectors. This device supports estimation of the cause of an anomaly detected in a target object for detecting an anomaly.

Abstract: One embodiment provides for a system to compute and distribute data for distributed training of a neural network, the system including first memory to store a first set of instructions including a machine learning framework; a fabric interface to enable transmission and receipt of data associated with the set of trainable machine learning parameters; a first set of general-purpose processor cores to execute the first set of instructions, the first set of instructions to provide a training workflow for computation of gradients for the trainable machine learning parameters and to communicate with a second set of instructions, the second set of instructions facilitate transmission and receipt of the gradients via the fabric interface; and a graphics processor to perform compute operations associated with the training workflow to generate the gradients for the trainable machine learning parameters.

Abstract: A computing system for simulating allocation of resources to a plurality of entities is disclosed. The computing system can be configured to input an entity profile that describes a preference and/or demand of a simulated entity into a reinforcement learning agent model and receive, as an output of the reinforcement learning agent model, an allocation output that describes a resource allocation for the simulated entity. The computing system can select one or more resources based on the resource allocation described by the allocation output and provide the resource(s) to an entity model that is configured to simulate a simulated response output that describes a response of the simulated entity. The computing system can receive, as an output of the entity model, the simulated response output and update a resource profile that describes the at least one resource and/or the entity profile based on the simulated response output.

Abstract: A mechanism is provided in a data processing system for training a computer implemented model. The mechanism determines an operation for which the computer implemented model is to be trained. The mechanism performs a statistical analysis of an enterprise dataset for an enterprise to generate one or more statistical distributions of cases and features correlated with the operation for which the computer implemented model is to be trained. The mechanism selects a subset of cases in the enterprise dataset for annotation based on the one or more statistical distributions of cases and features. The mechanism annotates the selected subset of cases to generate an annotated training dataset. The mechanism trains the computer implemented model, using the annotated training dataset, to perform the operation.

Abstract: Typical autonomous systems implement black-box models for tasks such as motion detection and triaging failure events, and as a result are unable to provide an explanation for its input features. An explainable framework may utilize one or more explainable white-box architectures. Explainable models allow for a new set of capabilities in industrial, commercial, and non-commercial applications, such as behavioral prediction and boundary settings, and therefore may provide additional safety mechanisms to be a part of the control loop of automated machinery, apparatus, and systems. An embodiment may provide a practical solution for the safe operation of automated machinery and systems based on the anticipation and prediction of consequences. The ability to guarantee a safe mode of operation in an autonomous system which may include machinery and robots which interact with human beings is a major unresolved problem which may be solved by an exemplary explainable framework.

Abstract: Example implementations relate to a failure policy. For example, in an implementation, storage device status data is encoded into storage device states. An action is chosen based on the storage device state according to a failure policy, where the failure policy prescribes, based on a probabilistic model, whether for a particular storage device state a corresponding action is to take no action or to initiate a failure mitigation procedure on a storage device. The failure policy is rewarded according to a timeliness of choosing to initiate the failure mitigation procedure relative to a failure of the storage device.

Abstract: Described is a system for computing conditional probabilities of random variables for Bayesian inference. The system implements a spiking neural network of neurons to compute the conditional probability of two random variables X and Y. The spiking neural network includes an increment path for a synaptic weight that is proportional to a product of the synaptic weight and a probability of X, a decrement path for the synaptic weight that is proportional to a probability of X, Y, and delay and spike timing dependent plasticity (STDP) parameters such that the synaptic weight increases and decreases with the same magnitude for a single firing event.

Abstract: The disclosed computer-implemented method for dynamically augmenting machine learning models based on contextual factors associated with execution environments may include (1) generating a base machine learning model and a supplemental set of machine learning models, (2) determining at least one contextual factor associated with an execution environment of a machine learning system that is configured to make predictions regarding a set of input data using at least the base machine learning model, (3) selecting, based on the contextual factor, a continuation set of machine learning models from the supplemental set of machine learning models, and (4) directing the machine learning system to utilize both the base machine learning model and the continuation set of machine learning models when making predictions regarding the set of input data. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: Systems and methods for dynamically managing figments are disclosed. A computer-implemented method includes: receiving, by a computing device, a question from a user; answering, by the computing device, the question using a first degree figment; classifying, by the computing device, the question based on topics; forwarding, by the computing device, the question to a set of second degree figments; receiving, by the computing device, answers to the question from the set of second degree figments; ranking, by the computing device, the answers received from the set of second degree figments; and providing, by the computing device, the ranked answers to the user.

Abstract: An apparatus to: analyze a data set to identify a candidate topic not in a set of topics; determine whether the prominence of the candidate topic within the data set meets a threshold; in response to meeting the threshold, retrieve a rate of increase in frequency of the candidate topic in online searches; in response to meeting a threshold rate of increase, retrieve the keyword most frequently used in online searches for the candidate topic, use the keyword to retrieve a supplemental data set, and analyze input data extracted from the supplemental data set to determine whether the candidate topic can change the accuracy of a forecast model; and in response to determining that the candidate topic can change the accuracy, add the candidate topic to the set of topics and replace the forecast model with a forecast model trained for the set of topics augmented with the candidate topic.

Abstract: Methods, systems, and computer program products for a collaborative cognition platform for creating and hosting social machines are provided herein. A computer-implemented method includes creating a social machine for collaborative tasks, wherein the social machine comprises (i) one or more human agents, (ii) one or more machine-based agents, (iii) an algorithm, and (iv) a set of rules prescribed for executing the collaborative tasks. The method also includes generating one or more collaborative resolutions for the collaborative tasks by executing, in an automated fashion, the collaborative tasks via implementation of the algorithm, wherein the algorithm facilitates, in accordance with the set of rules, systematic iterations of collaboration among (i) the one or more human agents and (ii) the one or more machine-based agents. Further, the method includes outputting the one or more collaborative resolutions to at least one user.

Abstract: A system and a method for a service engine providing distributed intelligent assistance to a user are described herein. The method comprising steps of receiving and displaying a user inquiry from the user, the user inquiry having a linguistic pattern including a verb; generating and displaying a follow up question based on the user inquiry; receiving and displaying a follow up answer from the user; and generating and displaying a response based on the user inquiry and the follow up answer.

Abstract: Disclosed are systems, methods, and non-transitory computer-readable media for structuring data in a knowledge graph. A data management system determines known concepts that are related to a data snippet. The data management system determines cosine similarity values indicating an intrinsic similarity between the data snippet and the known concepts, as well as pertinence values indicating a measure of topical similarity between the data snippet and the known concepts. The data management system determines, based on the cosine similarity values and the pertinence values, that the data snippet is related to a first known concept, and in response, assigns a concept identifier for the first known concept to the data snippet. Score indicating a strength of connection between the concepts added to the knowledge graph are determined and used to derive insights between the concepts.

Abstract: Various examples are described for using a movement sensor to detecting an activity of an infant. In an example, an activity classification system includes a sensor configured to measure the activity of an infant and an external monitor. The monitor receives, from a sensor, a time series of data comprising an inertial measurement for each a time period. The monitor determines, from the time series and by using a predictive model, an activity from a list of identified activities. Examples of identified activities are deep sleep, light sleep, sitting, awake, nursing, or bottle feeding.

Abstract: Methods and Systems for using reinforcement learning to optimize promotions. A promotion can be offered to a customer for a prepaid calling card using a reinforcement learning model with a sensitivity parameter. The reinforcement learning model can estimate a time period during which the customer will purchase the prepaid calling card. The customer's reaction to the promotion can be observed. A reward or a penalty can be collected based on the customer's reaction. The reinforcement learning model can be adapted based on the reward or the penalty to optimize the timing of the promotion by estimating a new time period during which the customer will purchase the prepaid calling card. The reward proxy and/or the penalty proxy can comprise frequency of usage.

Abstract: An approach is provided in which an information handling system uses historical time durations of deprecated resources to compute an increased probability window. The increased probability window corresponds to an increase in probability that a currently active resource is likely to be active at a future point in time. Next, the information handling system identifies a set of active resources that have active time durations within the increased probability window and, in turn, marks the set of resources as a set of forecasted active resources. In turn, the information handling system generates a resource cost forecast based on the set of forecasted active resources.

Abstract: A system and method for identifying and predicting subjective attributes for entities (e.g., media clips, movies, television shows, images, newspaper articles, blog entries, persons, organizations, commercial businesses, etc.) are disclosed. In one aspect, subjective attributes for a first media item are identified based on a reaction to the first media item, and relevancy scores for the subjective attributes with respect to the first media item are determined. A classifier is trained using (i) a training input comprising a set of features for the first media item, and a target output for the training input, the target output comprising the respective relevancy scores for the subjective attributes with respect to the first media item.

Abstract: Computer-implemented methods are provided for implementing training of a machine learning model in a heterogeneous processing system comprising a host computer operatively interconnected with an accelerator unit. The training includes a stochastic optimization process for optimizing a function of a training data matrix X, having data elements Xi,j with row coordinates i=1 to n and column coordinates j=1 to m, and a model vector w having elements wj. For successive batches of the training data, defined by respective subsets of one of the row coordinates and column coordinates, random numbers associated with respective coordinates in a current batch b are generated in the host computer and sent to the accelerator unit. In parallel with generating the random numbers for batch b, batch b is copied from the host computer to the accelerator unit.

Abstract: A method for inferring patterns in multi-dimensional image data comprises providing a recursive network of sub-networks with a parent feature node and at least two child feature nodes; wherein each sub-network is associated with a distinct subset of the space; configuring nodes of the sub-networks with posterior distribution component; receiving image data feature input at the final child feature nodes; propagating node activation through the network layer hierarchy in a manner consistent with node connections of sub-networks of the network and the posterior prediction of child nodes; and outputting parent feature node selection to an inferred output.

Abstract: A method and system for product recommendation. The method includes: defining, by a computing device, a hierarchical Bayesian model having a latent factor; training, by the computing device, the hierarchical Bayesian model using a plurality of training events to obtain a trained hierarchical Bayesian model, each event comprising feature of a product, brand of the product, feature of a user, and action of the user upon the product; predicting, by the computing device, a possibility a target user performing an action on a target product using the trained hierarchical Bayesian model; and providing product recommendation to the target user based on the possibility.