Abstract: This disclosure is related to determining an item push list for a user based on a reinforcement learning model. In one aspect, a method includes obtaining M first item lists that have been predetermined for a first user. Each first item list includes i?1 items. For each first item list, an ith state feature vector is obtained. The ith state feature vector includes a static feature and a dynamic feature. The ith state feature vector is provided as input to the reinforcement machine learning model. The reinforcement model outputs a weight vector including weights of sorting features. A sorting feature vector of each item in a candidate item set corresponding to the first item list is obtained. The sorting feature vector includes feature values of sorting features. M updated item lists are determined for the first item lists based on a score for each item in M candidate item sets.

Abstract: Triage of training data for acceleration of large-scale machine learning is provided. In various embodiments, training input from a set of training data is provided to an artificial neural network. The artificial neural network comprises a plurality of output neurons. Each output neuron corresponds to a class. From the artificial neural network, output values are determined at each of the plurality of output neurons. From the output values, a classification of the training input by the artificial neural network is determined. A confidence value of the classification is determined. Based on the confidence value, a probability of inclusion of the training input in subsequent training is determined. A subset of the set of training data is determined based on the probability. The artificial neural network is trained based on the subset.

Abstract: An automated way of learning archetypes which capture many aspects of entity behavior, and assigning entities to a mixture of archetypes, such that each entity is represented as a distribution across multiple archetypes. Given those representations in archetypes, anomalous behavior can be detected by finding misalignment with a plurality of entities archetype clustering within a hard segmentation. Extensions to sequence modeling are also discussed. Applications of this method include anti-money laundering (where the entities can be customers and accounts, as described extensively below), retail banking fraud detection, network security, and general anomaly detection.

Abstract: A method for using an artificial intelligence (AI) entity to interface with a customer relationship management (CRM) software platform online is provided. The method receives user input changes to the CRM software platform, by the AI entity comprising at least one processor and a memory element, wherein the AI entity is configured to perform chat-bot functionality; alters, by the at least one processor, contents of the CRM software platform associated with the user input changes, in response to the AI entity receiving the user input changes; continuously receives CRM data from the CRM software platform, by the at least one processor; receives, by the at least one processor, a user request for a subset of the CRM data; and in response to the user request, transmits the subset.

Abstract: Described herein is a system for automatically detecting invalid events in a distributed computing environment. The system for automatically detecting invalid events may include sub-systems and a learning engine. The learning engine may generate a rule set for each sub-system specifying circumstances under which an event is considered invalid specific to the sub-system using machine learning. Sub-systems may detect an invalid event being propagated through the distributed computing environment based on a set of rules specifying circumstances under which an event is considered invalid specific to the sub-system and/or metadata of the event.

Abstract: A method for determining an identity of a URL visited by a user from a vantage point in a network in which network traffic is encrypted includes determining a host to model, generating a list of URLs hosted by the host to model, repeatedly retrieving web resources referenced by the list of URLs hosted by the host to model and generating a network traffic signature upon each retrieval, generating a data feature for each of the generated network traffic signatures, and training, using the generated data features, a classifier corresponding to the host to model, wherein the classifier is configured to determine an identity of the visited URL from a signature of network traffic produced by the retrieval of a resource referenced by the visited URL.

Abstract: Relevance decay techniques are provided for time-based evaluation of machine learning applications and other classifiers. An exemplary method comprises obtaining time series measurement data; generating an input dataset comprising a plurality of records, wherein each record comprises features extracted from the time series measurement data, a target class corresponding to an event to be identified, and a time lag indicating a difference in time between a given extraction and the event to be identified; evaluating a plurality of classifiers during an evaluation phase using a portion of the input dataset and one or more predefined evaluation metrics weighted using a time-based relevance decay function based on the time lag; and selecting one or more of the classifiers to perform classification of the time series measurement data based on the predefined weighted evaluation metrics during a classification phase.

Abstract: Various techniques for temperature management during inductive energy transfer are disclosed. A transmitter device and/or a receiver device can be turned off during energy transfer based on the temperature of the transmitter device and/or of the receiver device.

Abstract: Backup battery systems for traffic cabinets that control traffic lights are provided herein. Backup battery systems include a controller operably coupled to 1 or more backup battery panels having rechargeable battery cells. Preferred systems can fit and operate entirely within the traffic cabinet. Monitoring and control of the backup system can be operable locally and remotely via internet cloud.

Abstract: Off-line deep neural network operations on client computing platforms may be enabled by cooperative machine learning across multiple client computing platforms and the cloud. A given client computing platform may include a client-side machine learning model configured to facilitate deep neural network operations on structured data. The operations may be performed by a client application installed on the given client computing platform. The client application may locally access the client-side machine learning model in order to perform the operations. Deep neural network operations on structured data may be performed at one or more servers. Sharing of model states may be facilitated between the cloud machine learning model and the client-side machine learning model. The cloud machine learning model may be improved, at one or more servers, based on usage of the application and user interactions with the given client computing platform.

Abstract: A method for predicting information propagation in a social network includes acquiring target information to be predicted, and acquiring influences of K clusters, where the target information is posted or forwarded by a first user at a first moment, and K is a positive integer; determining a role probability distribution of the first user, and determining a second user who has not propagated the target information, where the role probability distribution of the first user is used to indicate probabilities that the first user belongs separately to the K clusters; and determining, according to the influences of the K clusters and the role probability distribution of the first user, a probability that the second user forwards the target information from the first user. In the embodiments of the present application, propagation of target information in a social network can be predicted by using influences of K clusters.

Abstract: An individual antenna element is described. In one embodiment, the individual antenna element is a patch antenna which has an electrically small radiating element with a U-shaped slot on an antenna substrate, with a gain greater than 2 dBi with +/?45 degree coverage.

Abstract: A foldable display apparatus, a method of manufacturing the same, and a controlling method of the same are disclosed. The foldable display apparatus includes a substrate including a metal thin film and an insulating layer provided on the metal thin film, an organic light-emitting unit formed on the substrate and emitting light in an direction away from the substrate, and a thin film encapsulating layer for encapsulating the organic light-emitting unit. The foldable display apparatus may be folded in a direction such that the metal thin film is exposed.

Abstract: A specification of the process model is received. The process model includes a plurality of process components. A relationship between a first process component and another process component of the plurality of process components is determined using a predictive model. A process rule for the first process component is determined. The process rule specified a second process component to be executed. The process rule includes the relationship determined using the predictive model or a heuristic rule. The second process component is executed according to the process rule.

Abstract: Embodiments of an informatics platform where collected data can be normalized, integrated and mapped to a knowledge source, such as medical vocabulary systems are disclosed. One example of such a knowledge source is Unified Medical Language System (UMLS) which is a knowledge source for biomedical applications. Embodiments as depicted herein may provided a method to convert the desired information from UMLS into an ontology representation to allow for its use in conjunction with an informatics system.

Abstract: The disclosed computer-implemented method for identifying subject-matter experts may include (i) collecting, by the computing device, a plurality of electronic messages transmitted within an organization, (ii) creating a message graph for the organization, (iii) extracting a plurality of topics from the plurality of electronic messages transmitted within the organization, (iv) annotating the message graph by correlating each topic within the plurality of topics with each edge of the message graph that represents an electronic message related to the topic, and (v) identifying, based on an analysis of the annotated message graph, at least one vertex that represents an expert on at least one topic from the plurality of topics. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: An embodiment includes a method, comprising: pruning a layer of a neural network having multiple layers using a threshold; and repeating the pruning of the layer of the neural network using a different threshold until a pruning error of the pruned layer reaches a pruning error allowance.

Abstract: A method, system, and computer program product for obtaining a candidate event sequence that includes at least one event for achieving a goal, obtaining a reference event sequence, the candidate event sequence comprising at least one event that is not comprised in the reference event sequence, comparing an effectiveness of the candidate event sequence on the goal and an effectiveness of the reference event sequence on the goal, and identifying the candidate event sequence as an effective sequence for achieving the goal in response to the effectiveness of the candidate event sequence being better than the effectiveness of the reference event sequence.

Abstract: Disclosed is an online system that infers interests of unresolved users for whom the interests are not known. The online system determines certain features about the unresolved users, but does not have certain information about the users themselves (e.g., their interests), so instead infers these attributes based on the features of the user. The online system provides the features as input to a classifier trained to predict a particular interest, and the classifier outputs a prediction of whether the user has the corresponding interest. In one embodiment, the online system trains a classifier for various interest values by forming training sets for the interests using the features for users who are logged into the online system and hence have known interests.

Abstract: A user device can send, to a server, a request for a set of documents likely to be opened by a user, determine a client-suggested document to present to the user and a potential motive for the user to open the client-suggested document, receive a suggestion message from the server, the suggestion message including a set of documents likely to be opened by the user and potential motives for the user to open documents in the set of documents, and present, on a display of the user device, visual representations of the client-suggested document, the potential motive for the user to open the client-suggested document, multiple documents included in the set of documents, and the potential motives for the user to open the multiple documents in the set of documents.