Abstract: An RFID tag manufacturing method is provided that includes arranging a plurality of RFIC elements in a housing tool at a first arrangement density; and extracting an RFIC element group out of the plurality of RFIC elements arranged in the housing tool. Moreover, the extracted RFIC element group has a second arrangement density that is lower than the first arrangement density and that corresponds to an arrangement density of a plurality of antenna patterns arranged on an antenna substrate. The method includes disposing the RFIC element group onto the plurality of antenna patterns of the antenna substrate while maintaining the second arrangement density.

Abstract: Provided is a technique for preventing the occurrence of damage, such as breaking of an antenna, due to bending, in an IC tag in which an IC chip is protected with a reinforcing member. An IC tag according to an aspect of the present invention is an IC tag attachable to an article, and includes an inlay including an IC chip and an antenna electrically connected to the IC chip; at least one reinforcing member that is disposed on at least one face of the inlay so as to protect the IC chip; and at least one sheet-like cover member that is disposed on at least one face side of the inlay and that covers the reinforcing member. The cover member is not adhesively joined in a region around the reinforcing member and is adhesively joined to at least one of the inlay and the reinforcing member in another region.

Abstract: A grommet (20, 30) as part of a fastening element for a building envelope essentially comprises a head (21), a tip (22) and a sleeve (23) therebetween, wherein the head (21) is essentially constructed as an extensive washer (24), the central hole of which is adjoined by the tubular sleeve (23). The tip (22) narrows essentially in a conical or tapered manner to a smaller diameter. The grommet (20, 30) has a sensor arrangement (25), with at least one RFID transponder with antenna and a sensor operatively connected to the transponder.

Abstract: The invention relates to a UHF RFID transponder with an antenna and an RFID chip, wherein the antenna has a first dipole structure, a first inductive loop and an infeed point, and the first dipole, the inductive loop and the RFID chip are electrically connected to the infeed point. The purpose of the invention is to create an RFID transponder that can be tuned with reduced influence from the geometric tolerances involved in the production of the antenna. As a result, these RFID transponders are intended to be particularly suitable for attaching to glass panes, as the impedance of the antenna can be tuned to the impedance of the RFID chip with less influence from the thickness or material composition of the glass. This task is achieved in that the antenna (2) has a second inductive loop (8) and that this second inductive loop (8) is electrically connected to the infeed point (3) and connected in parallel to the first inductive loop (7).

Abstract: Processing raw data stored in an historian device for determining an amount of products passed through a process element in a process control environment is described. A count value is incremented by a counter at a rate at which products pass through the process element. The count value rolls over to zero when the count value reaches a rollover value R. An historian device periodically receives count value data points from the counter. A deadband value D is set in the historian device for distinguishing between rollovers, resets, and reversals. A client device queries the historian device for an amount of products passed through the process element for a timeframe. The historian device selects a set of count value data points from within the queried timeframe. The historian device determines, based on the selected data points and their quality, an amount of products passed through the process element.

Abstract: Race car settings (e.g., Formula 1 engine mix settings) are developed for particular racing goals such as faster lap time, better acceleration, less vehicle wear, etc., using pattern templates that are derived from historical racing scenarios. The historical scenarios provide data on racing settings, racing results, and racing conditions such as squad information, equipment information, and environmental information. A cognitive (deep question answering) system can select an initial pattern template based on current racing conditions, and present suggested vehicle settings to the user (driver) using the initial pattern template. The driver can select from different candidate values for various factors, which may lead to the presentation of additional suggestions or the use of additional pattern templates. The final settings map is created based on the employed pattern templates and the driver selections.

Abstract: In one embodiment, a method comprises determining that a membrane potential of a first neuron of a first neuron core exceeds a threshold; determining a first plurality of synapse cores that each store at least one synapse weight associated with the first neuron; and sending a spike message to the determined first plurality of synapse cores.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for environment simulation. In one aspect, a system comprises a recurrent neural network configured to, at each of a plurality of time steps, receive a preceding action for a preceding time step, update a preceding initial hidden state of the recurrent neural network from the preceding time step using the preceding action, update a preceding cell state of the recurrent neural network from the preceding time step using at least the initial hidden state for the time step, and determine a final hidden state for the time step using the cell state for the time step. The system further comprises a decoder neural network configured to receive the final hidden state for the time step and process the final hidden state to generate a predicted observation characterizing a predicted state of the environment at the time step.

Abstract: A computer-implemented method and system are described for learning a vector representation for unique identification codes. An example method may include generating a unique identification code list using one or more virtual interaction contexts, the unique identification code list being a list of unique identification codes, selecting a target unique identification code in the unique identification code list, and determining, from the unique identification code list, an input set of unique identification codes using the target unique identification code, the input set including the target unique identification code and one or more context unique identification codes. Some implementations may further include inputting the input set of unique identification codes into a semantic neural network model, the semantic neural network model including one or more weight matrices, and modifying the one or more weight matrices using the input set of unique identification codes.

Abstract: Embodiments of the invention relate to a multiplexed neural core circuit. One embodiment comprises a core circuit including a memory device that maintains neuronal attributes for multiple neurons. The memory device has multiple entries. Each entry maintains neuronal attributes for a corresponding neuron. The core circuit further comprises a controller for managing the memory device. In response to neuronal firing events targeting one of said neurons, the controller retrieves neuronal attributes for the target neuron from a corresponding entry of the memory device, and integrates said firing events based on the retrieved neuronal attributes to generate a firing event for the target neuron.

Abstract: A neuromorphic memory circuit including a programmable resistive memory element, an axon LIF pulse generator to generate an axon LIF pulse, a back propagation pulse generator to generate a back propagation pulse, a postsynaptic capacitor configured to build up a forward propagation LIF charge over time, and a presynaptic capacitor configured to build up a back propagation LIF charge over time. A first transistor activates a first discharge path from the postsynaptic capacitor through the programmable resistive memory element when the axon LIF pulse generator generates the axon LIF pulse. A second transistor activates a second discharge path from the presynaptic capacitor through the programmable resistive memory element when the back propagation pulse generator generates the back propagation pulse.

Abstract: A method of object recognition trains a convolutional neural network (CNN) with a set of training images, then classifies an image of an object using the trained CNN. A first layer of the CNN is trained by generating a set of first convolutional filters from eigenvectors produced from linear principal component analysis of patches of the training images. The training of each of multiple hidden layers CNN includes generating a set of convolutional filters from a selected subset of eigenvectors produced from linear principal component analysis of patches of an affinity matrix constructed using a set of prior convolutional filters from a prior layer of the CNN, where the affinity matrix represents correlations of feature vectors associated with the prior layer. The last layer of the CNN is trained with a regular classifier by error back-propagation using the training images and labels associated with the training images.

Abstract: In one aspect, the inventions include a system for control of an entertainment state system. First, an application plane layer is adapted to receive instructions regarding operation of the entertainment state system. Preferably, the application plane layer is coupled to an application plane layer interface. Second, a control plane layer includes an adaptive control unit, such as a cognitive computing unit, an artificial intelligence unit or a machine-learning unit. Third, a data plane layer includes an input interface to receive data input from one or more data sources. An adaptive control unit is trained at least in part on analyzing the behavioral responses of users in response to content provided on one or more displays to the users as sensed by one or more sensors.

Abstract: Feature selection methods and processes that facilitate reduction of model components available for iterative modeling. It has been discovered that methods of eliminating model components that do not meaningfully contribute to a solution can be preliminarily discovered and discarded, thereby dramatically decreasing computational requirements in iterative programming techniques. This development unlocks the ability of iterative modeling to be used to solve complex problems that, in the past, would have required computation time on orders of magnitude too great to be useful.

Abstract: A method for training a deep learning network includes defining a loss function corresponding to the network. Training samples are received and current parameter values are set to initial parameter values. Then, a computing platform is used to perform an optimization method which iteratively minimizes the loss function. Each iteration comprises the following steps. An eigCG solver is applied to determine a descent direction by minimizing a local approximated quadratic model of the loss function with respect to current parameter values and the training dataset. An approximate leftmost eigenvector and eigenvalue is determined while solving the Newton system. The approximate leftmost eigenvector is used as negative curvature direction to prevent the optimization method from converging to saddle points. Curvilinear and adaptive line-searches are used to guide the optimization method to a local minimum. At the end of the iteration, the current parameter values are updated based on the descent direction.

Abstract: An apparatus for executing backpropagation of an artificial neural network comprises an instruction caching unit, a controller unit, a direct memory access unit, an interconnection unit, a master computation module, and multiple slave computation modules. For each layer in a multilayer neural network, weighted summation may be performed on input gradient vectors to calculate an output gradient vector of this layer. The output gradient vector may be multiplied by a derivative value of a next-layer activation function on which forward operation is performed, so that a next-layer input gradient vector can be obtained. The input gradient vector may be multiplied by an input neuron counterpoint in forward operation to obtain the gradient of a weight value of this layer, and the weight value of this layer can be updated according to the gradient of the obtained weight value of this layer.

Abstract: An apparatus for executing backpropagation of an artificial neural network comprises an instruction caching unit, a controller unit, a direct memory access unit, an interconnection unit, a master computation module, and multiple slave computation modules. For each layer in a multilayer neural network, weighted summation may be performed on input gradient vectors to calculate an output gradient vector of this layer. The output gradient vector may be multiplied by a derivative value of a next-layer activation function on which forward operation is performed, so that a next-layer input gradient vector can be obtained. The input gradient vector may be multiplied by an input neuron counterpoint in forward operation to obtain the gradient of a weight value of this layer, and the weight value of this layer can be updated according to the gradient of the obtained weight value of this layer.

Abstract: System, methods, and other embodiments described herein relate to improving the generation of realistic images. In one embodiment, a method includes acquiring a synthetic image including identified labels of simulated components within the synthetic image. The synthetic image is a simulated visualization and the identified labels distinguish between the components within the synthetic image. The method includes computing, from the simulated components, translated components that visually approximate real instances of the simulated components by using a generative module comprised of neural networks that are configured to separately generate the translated components. The method includes blending the translated components together to produce a new image from the simulated components of the synthetic image.

Abstract: Techniques are provided for determining labels associated with first and second candidate code and whether those labels are compatible. When the first candidate code and the second candidate code are compatible, third candidate code based is determined on the those two. When the third candidate code meets exit criteria the third candidate code is provided as evolved code. Some embodiments also include causing execution of the evolved code.

Abstract: A mechanism is provided in a data processing system for automatically generating question and answer pairs for training a question answering system for a given domain. The mechanism receives user input of question text for a question to be submitted to a question answering system in a user interface. The mechanism determines a question strength score for the question text. The question strength score represents a likelihood the question text will result in a correct answer with high confidence. The mechanism presents a graphical representation of the question strength score in the user interface.

Abstract: A framework and interface for invoking and assimilating external algorithms and interacting with said algorithms in-session and real-time are described herein. An example embodiment also includes reproducible, updatable nodes that can be leveraged for data-driven analysis whereby the data itself can direct the algorithm choice, variables, and presentation leading to iteration and optimization in an analysis workflow. With example embodiments, an entire discovery or diagnosis process may be executed on a particular data set, thereby divorcing the discovery or diagnosis process from a specific data set such that the same discovery or diagnosis process, phenotype identification, and visualizations may be repeated on future experiments, published, validated, or shared with another investigator.

Abstract: A method and system for identifying and prioritizing business useful insights from hidden patterns. This invention relates to data mining techniques and more particularly to identify and prioritize insights from a plurality of insights present in a large set of data. Insight exploration is a method and system that enables the user to generate actionable insights, prioritize them for a given data. This falls broadly within the field of data mining. The primary achievement of this invention is to take a rule in if-then format and then systematically process them to identify actionable information from them. In that process, the system automatically prioritizes the rules, generates other rules and analyzes the path that leads to desired behavioral changes.

Abstract: Approaches are provided for answering an inquiry of a cognitive distributed network. An approach includes receiving the inquiry at the cognitive distributed network. The approach further includes determining a classification for the inquiry based on natural language of the inquiry. The approach further includes classifying the inquiry as a single question class. The approach further includes determining, by at least one computing device, a type of introspection to be used by the cognitive distributed network on the inquiry. The approach further includes generating an answer to the inquiry based on the determined type of introspection.

Abstract: A system and computer-readable medium for performing a cognitive learning operation comprising monitoring a user interaction of a user; generating user interaction data based upon the user interaction; receiving data from a plurality of data sources; processing the user interaction data and the data from the plurality of data sources to perform a cognitive learning operation, the processing being performed via a cognitive inference and learning system, the cognitive learning operation comprising analyzing the user interaction data, the cognitive learning operation generating a cognitive learning result based upon the user interaction data; and, associating a cognitive profile with the user based the cognitive learning result.

Abstract: A cognitive learning method comprising: monitoring a user interaction of a user; generating user interaction data based upon the user interaction; receiving data from a plurality of data sources; processing the user interaction data and the data from the plurality of data sources to perform a cognitive learning operation, the processing being performed via a cognitive inference and learning system, the cognitive learning operation comprising analyzing the user interaction data, the cognitive learning operation generating a cognitive learning result based upon the user interaction data; and, associating a cognitive profile with the user based the cognitive learning result.

Abstract: In order to facilitate the entity resolution and entity activity tracking and indexing, systems and methods include receiving first source records from a first database and second source records from a record database. A candidate set of second source records is determined by a heuristic search in the set of second source records. A candidate pair feature vector associated with each candidate pair of first and second source records is generated. An entity matching machine learning model predicts matching first source records for each candidate second source record based on the respective candidate pair feature vector. An aggregate quantity associated with the matching first source records is aggregated from a quantity associated with each first source record, and a quantity index for each candidate second source record is determined based the aggregate quantities. Each quantity index is displayed to a user.

Abstract: In one implementation, a method includes obtaining time series data. The time serious data includes a plurality of network utilization measurements. The plurality of network utilization measurements is indicative of a plurality of utilizations of one or more resources of a network resource at a plurality of times. The method also includes determining whether the time series data comprises a plurality of segments. Each segment of the plurality of segments is associated with a separate regression model and each segment includes a portion of the time series data. The method further includes identifying a current segment from the time series data when the time series data comprises the plurality of segments. The method further includes determining an estimated network utilization based on a current regression model associated with the current segment.

Abstract: A method estimates similarities in data streams. A data source receives input vectors from users. A sketch feature generator converts the input vectors into sketch feature vectors corresponding to the users, wherein each sketch feature vector represents data and meta-information from each user received in a most recent sample period. A similarity comparator compares each sketch feature vector against other sketch feature vectors to calculate similarity probabilities between the users. A processor running a decision loop determines a prediction result for at least one user based on the similarity probabilities.

Abstract: Provided is a failure prediction method using a failure prediction apparatus, including: receiving time-series data about measured performance parameters from a device under test; encoding the time-series data with a plurality of symbols corresponding to a predetermined range; calculating a transition probabilities between the symbols of the encoded time-series data, and generating a transition matrix according to the transition probabilities; calculating an abnormal indicator, which is a difference between the transition matrix and a pre-stored database, and an increased value of the abnormal indicator; and comparing the increased value of the abnormal indicator and a predetermined threshold value, and if the increased value of the abnormal indicator is greater than the predetermined threshold value, predicting that failure of the device under test is to occur.

Abstract: A computer-implemented method is provided for hiding overheads on a parallel computing platform. The computer-implemented method includes loading a first kernel overhead and a second kernel overhead in a queue of a second thread, and loading a dummy kernel overhead between the first and second kernel overheads in the queue of second thread. The computer-implemented method further includes loading a waiting process in the queue of a first thread, the waiting process remaining active until a previous kernel of the first and second kernel overheads ends. The computer-implemented method furthers include allocating memory copy overheads related to the previous kernel in the queue of the first thread and allocating a stop process in the queue of the first thread, the stop process configured to stop a dummy kernel, the dummy kernel related to the dummy kernel overhead.

Abstract: Described herein are methods, systems, and media for generating a quantum-ready or quantum-enabled software development kit (SDK) for a quantum computing system. Such methods may comprise accepting user input from an application at an application interface, which application is executed on a digital computer, and implementing one or more algorithms, at an algorithms layer, that may be solved heuristically or exactly depending on the requirements of the user input. The one or more algorithms may abstract away a complexity of the application; transforming the one or more algorithms from the application space into the one or more instructions in polynomial unconstrained binary optimization (PUBO) form. The one or more instruction may be executed in PUBO form at the common interface solver layer.

Abstract: A method includes depositing a first layer on a portion of a first surface of a quantum hardware, the portion of the first surface comprising a set of wirebonds. The method further includes coupling the set of wirebonds to the first layer. The method further includes removing the first layer and the set of wirebonds from the first surface of the quantum hardware. In an embodiment, the first layer is an inert polymer in solution.

Abstract: Among other things, an apparatus comprises quantum units; and couplers among the quantum units. Each coupler is configured to couple a pair of quantum units according to a quantum Hamiltonian characterization of the quantum by the coupler.

Abstract: Systems and techniques are provided for template exploration in a large-scale machine learning system. A method may include obtaining multiple base templates, each base template comprising multiple features. A template performance score may be obtained for each base template and a first base template may be selected from among the multiple base templates based on the template performance score of the first base template. Multiple cross-templates may be constructed by generating a cross-template of the selected first base template and each of the multiple base templates. Performance of a machine learning model may be tested based on each cross-template to generate a cross-template performance score for each of the cross-templates. A first cross-template may be selected from among the multiple cross-templates based on the cross-template performance score of the cross-template. Accordingly, the first cross-template may be added to the machine learning model.

Abstract: This document discloses a system and method for consolidating threat intelligence data for a computer and its related networks. Massive volumes of raw threat intelligence data are collected from a plurality of sources and are partitioned into a common format for cluster analysis whereby the clustering of the data is done using unsupervised machine learning algorithms. The resulting organized threat intelligence data subsequently undergoes a weighted asset based threat severity level correlation process. All the intermediary network vulnerabilities of a particular computer network are utilized as the critical consolidation parameters of this process. The final processed intelligence data gathered through this high speed automated process is then formatted into predefined formats prior to transmission to third parties.

Abstract: This disclosure provides a method and system to perform data integrity checks in a data warehouse (DWH) feed using machine-learning (ML) processes. According to an exemplary method, a ML integrity check is performed on received data which has been extracted from a plurality of source files, and after ML processes validate the extracted data, the validated data is transformed and loaded to a DWH.

Abstract: A system for processing social media data includes a platform with a social media acquisition module configurable to collect a plurality of social media statements. The platform further includes an analysis engine configurable to analyze the plurality of social media statements according to a first sentiment model to generate first analytics data. The analysis engine is configurable to present the first analytics data to a client user, including a display of a sentiment value for each of the social media statements. The platform further includes a feedback queue configurable to receive feedback from the client user on at least a portion of the sentiment values; a model modification module configurable to modify the first sentiment model based on the feedback to result in a modified sentiment model; and a database configurable to store the modified sentiment model as a personalized sentiment model for the client user.

Abstract: A determination that a machine learning data set is to be shuffled is made. Tokens corresponding to the individual observation records are generated based on respective identifiers of the records' storage objects and record key values. Respective representative values are derived from the tokens. The observation records are rearranged based on a result of sorting the representative values and provided to a shuffle result destination.

Abstract: Classifier generation methods are described in which features used in classification (e.g., mass spectral peaks) are selected, or deselected using bagged filtering. A development sample set is split into two subsets, one of which is used as a training set the other of which is set aside. We define a classifier (e.g., K-nearest neighbor, decision tree, margin-based classifier or other) using the training subset and at least one of the features (or subsets of two or more features in combination). We apply the classifier to a subset of samples. A filter is applied to the performance of the classifier on the sample subset and the at least one feature is added to a “filtered feature list” if the classifier performance passes the filter. We do this for many different realizations of the separation of the development sample set into two subsets, and, for each realization, different features or sets of features in combination.

Abstract: A system for providing adaptive metric pruning includes a processor; a memory; and one or more modules stored in the memory and executable by a processor to perform operations including: receive, by a collector, metadata associated with the business transaction running in a monitored environment; train a machine learning system by providing training data and an anticipated result for the training data to the machine learning system to generate rules for retaining given metadata; predict a retention requirement for the received metadata by providing the received metadata to the machine learning system to apply the generated rules and generate a result for retaining the received metadata; and provide a user interface to display the generated result for retaining the received metadata including a recommendation on how to retain the received metadata according to the result for retaining the received metadata.

Abstract: The disclosed systems and methods include pre-calculation, per object, of object feature bin values, for identifying close matches between objects, such as text documents, that have numerous weighted features, such as specific-length word sequences. Predetermined feature weights get scaled with two or more selected adjacent scaling factors, and randomly rounded. The expanded set of weighted features of an object gets min-hashed into a predetermined number of feature bins. For each feature that qualifies to be inserted by min-hashing into a particular feature bin, and across successive feature bins, the expanded set of weighted features get min-hashed and circularly smeared into the predetermined number of feature bins. Completed pre-calculated sets of feature bin values for each scaling of the object, together with the scaling factor, are stored for use in comparing sampled features of the object with sampled features of other objects by calculating an estimated Jaccard similarity index.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media for performing machine learning tasks. One method includes receiving (i) a model input, and (ii) data identifying a first machine learning task to be performed on the model input to generate a first type of model output for the model input; augmenting the model input with an identifier for the first machine learning task to generate an augmented model input; and processing the augmented model input using a machine learning model, wherein the machine learning model has been trained on training data to perform a plurality of machine learning tasks including the first machine learning task, and wherein the machine learning model has been configured through training to process the augmented model input to generate a machine learning model output of the first type for the model input.

Abstract: In accordance with disclosed embodiments, there are provided systems, methods, and apparatuses for implementing machine learning model training and deployment with a rollback mechanism within a computing environment.

Abstract: Controlling circumferential variability in a blast furnace may include generating a predictive model that sets up a relationship between a standard deviation of a selected state variable, state variables and one or more control variables in blast furnace operation for predicting the standard deviation. A number of circumferential sections of the blast furnace is defined, and the predictive model associated with the selected state variable for each of the circumferential sections is trained based on process data of the blast furnace. A plurality trained predictive models is generated associated with different circumferential sections and different selected state variables. One or more future control variable set points that minimize a sum of the plurality of predictive models, is determined. One or more future control variable set points is transmitted to a control system to control the blast furnace operation.

Abstract: Embodiments are described for transforming data relatively homogeneous data segments for training or otherwise configuring automated modeling systems. For example, a computing system selects, from standardized data, geographic data objects associated with a completed transaction value greater than the threshold number. The computing system clusters these geographic data objects and combines completed transactions associated with the selected geographic data objects into segments. The computing system performs a similarity analysis for pairs of the geographic data objects by comparing numerically valued attributes of the geographic data objects. The computing system also selects, from the standardized data, remaining completed transactions that are not grouped into segments and combines, based on the similarity analysis and the threshold number, these property transactions into segments such that each completed transaction has been assigned to one or more of the relatively homogeneous segments.

Abstract: Historical data used to train machine learning algorithms can have thousands of records with hundreds of fields, and inevitably includes faulty data that affects the accuracy and utility of a primary model machine learning algorithm. To improve dataset integrity it is segregated into a clean dataset having no invalid data values and a faulty dataset having the invalid data values. The clean dataset is used to produce a secondary model machine learning algorithm trained to generate from plural complete data records a replacement value for a single invalid data value in a data record, and a tertiary model machine learning clustering algorithm trained to generate from plural complete data records replacement values for multiple invalid data values. Substituting the replacement data values for invalid data values in the faulty dataset creates augmented training data which is combined with clean data to train a more accurate and useful primary model.

Abstract: A server of a transportation network determines characteristics of trips provided to users, as well as the usage of the services by users. Using the determined characteristics, the server trains a model that, for a given set of trips for a user, estimates a degree of likely user dissatisfaction with the trips. Based on the estimated degree of user dissatisfaction, the system can estimate user dissatisfaction in real time, directly after completion of a trip, and can take remedial actions should the user be estimated to be likely dissatisfied.

Abstract: A system is disclosed for managing waste services. The system may include a waste detector mounted onboard a service vehicle and configured to capture image data of a receptacle encountered by the service vehicle. The system may also include a display, and a controller in communication with the waste detector and the display. The controller may be configured to recognize the receptacle within the image data, and to determine a fill status of the receptacle based on the image data. The controller may also be configured to selectively determine a customer recommendation for a different sized receptacle based on the fill status, and to cause a dispatch to provide the different sized receptacle based on the customer recommendation.

Abstract: Methods, apparatus, and computer program products for determining software complexity. A plurality of versions of a software module whose complexity is to be determined are compressed. Lengths of the compressed versions are compared, one with another, to provide complexity metrics.

Abstract: Aspects of the technology described herein provide a more efficient user interface by providing suggestions that are tailored to a specific user's interests. The suggestions may be provided by a personal assistant or some other application running on a user's computing device. A goal of the technology described herein is to provide relevant suggestions when the user can and actually wants to use them. The suggestions are designed to provide information or services the user wants to use.