Abstract: A method includes provisioning a set of training data sets, each of the training data sets respectively including an acquisition data set; generating a first medical image for each of the training data sets of at least one first subset of the set of training data sets using the image reconstruction algorithm based on a respective acquisition data set; determining an image processing result for each of the respective first medical images using an image processing algorithm based on the respective first medical image; determining image processing information for each of the respective first medical images relating to a quality of the respective image processing result based on the respective image processing result; and optimizing the image reconstruction algorithm based on a first machine learning algorithm, the at least one first subset of the set of training data sets and the image processing information for the respective first medical images.

Abstract: Disclosed are techniques for generating a personalized flow for a software delivery model. These techniques identify a plurality of business clusters and a plurality of expense categories for the plurality of business clusters and natural language description about a first characteristic of a specific user from an electronic tax return in an electronic tax return preparation and filing software application delivery model. One or more business clusters may be determined from the plurality of business clusters for the specific user based in part or in whole upon a user description vector of at least a part of the description. One or more expense categories may be identified for the electronic tax return of the specific user based in part or in whole upon the one or more business clusters.

Abstract: Technical solutions are described for predicting linepack delays. An example method includes receiving temporal sensor measurements of a first fluid-delivery pipeline network and generating a causality graph of the first fluid-delivery pipeline network. The method also includes determining a topological network of the stations based on the causality graph, where the topological network identifies a temporal delay between a pair of stations. The method also includes generating a temporal delay prediction model based on the topological network and predicting the linepack delays of a second fluid-delivery pipeline network based on the temporal delay prediction model, where a compressor station of the second fluid-delivery pipeline network compresses fluid based on the predicted linepack delays to maintain a predetermined pressure.

Abstract: Aspects of the present disclosure describe techniques for training a convolutional neural network using an inconsistent stochastic gradient descent (ISGD) algorithm. Training effort for training batches used by the ISGD algorithm are dynamically adjusted according to a determined loss for a given training batch which are classified into two sub states—well-trained or under-trained. The ISGD algorithm provides more iterations for under-trained batches while reducing iterations for well-trained ones.

Abstract: Devices, systems, and method are provided for operating a store. One method includes sampling a shopping environment using one or more sensors to produce features relating to a state of activity of the store. The method further includes processing output of at least one sensor through a feature extractor to extract one or more additional features relating to the state of activity of the retail store. Then, processing at least part of the produced features and at least part of the extracted additional features using a processing entity associated with the store and a machine learning model to generate one or more labels relating to the state of activity of the retail store. In further embodiments, items in the store are physically taken into possession and automatically added to a user's electronic shopping cart to enable processing of a cashier-less transaction.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating speech from text. One of the systems includes one or more computers and one or more storage devices storing instructions that when executed by one or more computers cause the one or more computers to implement: a sequence-to-sequence recurrent neural network configured to: receive a sequence of characters in a particular natural language, and process the sequence of characters to generate a spectrogram of a verbal utterance of the sequence of characters in the particular natural language; and a subsystem configured to: receive the sequence of characters in the particular natural language, and provide the sequence of characters as input to the sequence-to-sequence recurrent neural network to obtain as output the spectrogram of the verbal utterance of the sequence of characters in the particular natural language.

Abstract: A method, device, and computer program storage product for generating a query to extract clinical features into a set of electronic medical record (EMR) tables based on clinical knowledge. A knowledge tree is constructed according to a set of clinical knowledge data. An EMR graph corresponding to a set of EMR tables is obtained. The EMR graph comprises at set of table nodes and a set of attribute nodes. The set of table nodes and the set of attribute nodes represent a structure of each EMR table in the set of EMR tables and a reference relationship among attributes of set of EMR tables. A plurality of sub-queries is generated based on the knowledge tree and the EMR graph. At least one query is generated by combining the plurality of sub-queries according to the knowledge tree.

Abstract: A method of analyzing data exchange of at least one device includes feeding a plurality of data exchanged by the at least one device to a system for data exchange analysis that includes a deep learning algorithm. The deep learning algorithm includes at least an input layer, an output layer of the same size as the input layer, and hidden layers. Neurons of the hidden layers receive recurrently, at each time t, only a subset of the data exchanged by the at least one device up to time t, the subset of data comprising current data from time t and only a fraction of past data from time tpast to time t, with tpast<t. The method includes attempting to reconstruct, at the output layer, at each time t, data received at the input layer. The reconstructed data is compared with at least part of the plurality of data. In indication is provided on one or more anomalies in the data, based on at least the comparison.

Abstract: A method and system for generating synthetic form image involves obtaining a multitude of field value data and associated field labels for a chosen type of form document from an electronic data source, classifying the multitude of field value data into a multitude of data categories, where the multitude of data categories, learning statistical data distributions for categorical and numerical data types using the classified categorical and numerical data, and sampling data elements randomly using the learned data distributions to generate synthetic data for categorical and numerical data. The method also involves assembling the synthetic data for the multitude of data categories with the associated field labels to generate a labeled synthetic textual data set, rendering the labeled synthetic textual data set over a structured form layout image to produce a synthetic form image, and storing the synthetic form image and the labeled synthetic textual data set.

Abstract: An approach is provided that receives a set of user information pertaining to a user. The received set of information is encoded into a neural network and the neural network is trained using the encoded user information. As an output of the trained neural network, passwords corresponding to the user are generated.

Abstract: A data analysis supporting apparatus according to an embodiment comprises a variable-type determining device, a variable-variation generating device, and a variable-contribution rate determining device. The variable-type determining device determines the type of each variable. The variable-variation generating device generates variations of each variable in accordance with the variable type determined and with a variable variation rule that defines a method of generating the variations of each variable of any type. The variable-contribution rate determining device calculates the rate at which the variations and analysis data contribute to the objective variable, and determines, from the rate, whether the variations should be used or deleted.

Abstract: Devices, systems, and method are provided for identifying actions in a retail store. One method includes sampling a shopping environment using one or more sensors. The method includes receiving output of the sampling as feature inputs to a machine learning classifier model to derive one or more labels characterizing a state of an item. At least one processing entity associated with the retail store detects the state of the item to change from one as item taken to one as item returned, and sensor data from said one or more sensors characterizing the item as having been returned to a wrong location in the retail store. In further embodiments, items taken are added to a user's electronic shopping cart to enable processing of a cashier-less transaction.

Abstract: A processing system for processing data defining behaviour of a physical system, the processing system being configured to operate on: (i) a plurality of data definition objects, each data definition object defining a method for computing one or more output result values for that object in dependence on one or more input values for that object; and (ii) a dependency schema defining one or more dependencies among the data definition objects, each dependency being such that a first data object is dependent on a second data object when the first data object takes as a required input value an output value of the second data object; the data processing system being capable of raising each data definition object to a computed state by executing the method defined in the respective data definition object to thereby compute one or more output values for that data definition object; and being configured to receive from a user an instruction indicating a request to raise a specific data definition object to a computed

Abstract: The present disclosure is directed to parallelization of artificial neural network processing by conditionally synchronizing, among multiple computer processors, either the input or output of individual operations, and by conditionally using either rows or columns of certain matrices used in the operations. The conditional processing may depend upon the relative sizes of the input and output of the specific operations to be performed. For example, if a current layer matrix of values is larger than a next layer matrix of values to be computed, then rows of a weight matrix may be used by the computer processors to compute the next layer matrix. If the current layer matrix is smaller than the next layer matrix, then columns of the weight matrix may be used by the computer processors to compute the next layer matrix.

Abstract: An apparatus, device, methods, computer program product, and system are described that provide a graphical illustration of a first possible outcome of a use of a treatment parameter with respect to at least one body portion, the first possible outcome based on at least one dataset associated with at least one predictive basis, and apply a filter criteria to the at least one dataset to provide a modified graphical illustration of a second possible outcome of the use of the treatment parameter.

Abstract: A three-dimensional stacked integrated circuit (3D SIC) for implementing an artificial neural network (ANN) having a non-volatile memory die including an array of non-volatile memory partitions, wherein each partition of the array of non-volatile memory partitions is configured to store first parameters of a set of neurons. The 3D SIC also has a volatile memory die including an array of volatile memory partitions, wherein each partition of the array of volatile memory partitions is configured to store second parameters of the set of neurons. The 3D SIC also has a processing logic die including an array of processing logic partitions. Each one of the partitions of the array of processing logic partitions is configured to receive input data and process the input data according to the set of neurons to generate output data.

Abstract: A device to enhance and present a medical image using a corrective mechanism is described. An image analysis application executed by the device captures a digital copy of the medical image displayed on a display device. A flawed photography effect associated with the digital copy is identified by processing the digital copy. Next, the digital copy is enhanced based on the flawed photography effect. Furthermore, the enhanced digital copy can be processed with an artificial intelligence mechanism to generate an annotation. The annotation is associated with a cancer identification. In addition, the enhanced digital copy and the annotation are displayed.

Abstract: A mobile robot is configured for operation in a commercial or industrial setting, such as an office building or retail store. The robot can patrol one or more routes within a building, and can detect violations of security policies by objects, building infrastructure and security systems, or individuals. In response to the detected violations, the robot can perform one or more security operations. The robot can include a removable fabric panel, enabling sensors within the robot body to capture signals that propagate through the fabric. In addition, the robot can scan RFID tags of objects within an area, for instance coupled to store inventory. Likewise, the robot can generate or update one or more semantic maps for use by the robot in navigating an area and for measuring compliance with security policies.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating output sequences from input sequences. One of the methods includes obtaining an input sequence having a first number of inputs arranged according to an input order; processing each input in the input sequence using an encoder recurrent neural network to generate a respective encoder hidden state for each input in the input sequence; and generating an output sequence having a second number of outputs arranged according to an output order, each output in the output sequence being selected from the inputs in the input sequence, comprising, for each position in the output order: generating a softmax output for the position using the encoder hidden states that is a pointer into the input sequence; and selecting an input from the input sequence as the output at the position using the softmax output.

Abstract: The present disclosure provides methods for applying artificial neural networks to flow cytometry data generated from biological samples to diagnose and characterize cancer in a subject. The disclosure also provides methods of training, testing, and validating artificial neural networks.

Abstract: Embodiments of an electronic fraud analysis platform system are disclosed which may be used to analyze tax returns for potential fraud. Analysis of tax return data using the tax return analysis platform computing systems and methods discussed herein may provide insight into whether a tax return may be fraudulent based on, for example, an initial screening component configured to filter tax returns which appear fraudulent due to missing or inaccurate information provided with the return; a device activity analysis component configured to identify whether a device used to submit a tax return or to provide further authentication information needed to complete processing of the return may have been used in other fraudulent activities; and a knowledge-based authentication component configured to identify potential fraudsters using dynamically generated questions for which fraudsters typically do not know the answers.

Abstract: Systems and methods are disclosed for operating a Restricted Boltzmann Machine (RBM) by determining a corrected energy function of high-order semi-RBMs (hs-RBMs) without self-interaction; performing distributed pre-training of the hs-RBM; adjusting weights of the hs-RBM using contrastive divergence; generating predictions by Gibbs Sampling or by determining conditional probabilities with hidden units integrated out; and generating predictions.

Abstract: An information processing system includes a state data acquisition device configured to acquire state data indicating an operation state of a plant, a first classification device configured to generate a determination line used to classify the state data, a mirror image data generation device configured to generate mirror image data obtained by mirroring the acquired state data using the generated determination line, and a second classification device configured to generate a determination model on the basis of the state data and the generated mirror image data.

Abstract: Discriminative pretraining technique embodiments are presented that pretrain the hidden layers of a Deep Neural Network (DNN). In general, a one-hidden-layer neural network is trained first using labels discriminatively with error back-propagation (BP). Then, after discarding an output layer in the previous one-hidden-layer neural network, another randomly initialized hidden layer is added on top of the previously trained hidden layer along with a new output layer that represents the targets for classification or recognition. The resulting multiple-hidden-layer DNN is then discriminatively trained using the same strategy, and so on until the desired number of hidden layers is reached. This produces a pretrained DNN. The discriminative pretraining technique embodiments have the advantage of bringing the DNN layer weights close to a good local optimum, while still leaving them in a range with a high gradient so that they can be fine-tuned effectively.

Abstract: Aspects of group signatures with probabilistic revocation are described. In one example employing these aspects, at least one computing device can map an alias token to an alias code comprising a plurality of alias code segments. Each of the alias code segments is based at least in part on a set of orthogonal codes. Also, each of the alias code segments corresponds to a segment of the alias token. A revocation code is based at least in part on a plurality of revoked alias codes. One of the alias code segments and a corresponding segment of the revocation code can be utilized to determine a revocation status of the alias token.

Abstract: Storing and/or accessing data in a transactional database, with use of the following technique: (i) selecting a set of base tables in a transactional database; and (ii) creating a factless materialized query table, having maximum sparsity, for the set of base tables. The set of base tables includes at least two base tables. The set of base tables includes a set of keys including at least two distinct primary keys. The factless materialized query table includes one record associated with each record in the set of base tables. Each record in the materialized query table includes a value for every primary key in the set of base tables.

Abstract: A processing node in a temporal memory system includes a spatial pooler and a sequence processor. The spatial pooler generates a spatial pooler signal representing similarity between received spatial patterns in an input signal and stored co-occurrence patterns. The spatial pooler signal is represented by a combination of elements that are active or inactive. Each co-occurrence pattern is mapped to different subsets of elements of an input signal. The spatial pooler signal is fed to a sequence processor receiving and processed to learn, recognize and predict temporal sequences in the input signal. The sequence processor includes one or more columns, each column including one or more cells. A subset of columns may be selected by the spatial pooler signal, causing one or more cells in these columns to activate.

Abstract: A computer-implemented method for determining modeled roof age of a structure is provided. Such methods comprise identifying a target structure for which an estimated roof age is desired, receiving data related to the target structure in computer-readable form, and inputting the data related to the target structure into a generalized linear model to determine a Per-Property Modeled Roof Age of the target structure. The modeled roof age preferably accounts for (i) the height of the trees, (ii) the proximity of the trees to the target structure, and (iii) the weather conditions for the structure's geographic location. Such methods can be applied in practice by substituting a typically inaccurate reported roof age with a more realistic modeled roof age in making homeowners insurance decisions.

Abstract: Aspects of the embodiments are directed to systems, methods, and devices for generating a design of experiments (DOE) matrix, the DOE matrix comprising a set of possible combinations of values for a plurality of electrical parameters; iteratively applying each combination of values for the plurality of electrical parameters to one or more memory pins; determining a margin response for each combination of values; generating a prediction function based on a correlation of the margin response and each combination of values; and optimizing the plurality of electrical parameters based on the prediction function.

Abstract: A method for real-time control of thermal loads in a data center includes storing, in a database, a heat generation data associated with an electronic equipment located in a predetermined area of the data center, transmitting the heat generation data to a data analysis system in communication with a tunable temperature controller, the heat generation data provides a real-time input of an amount of heat generated in the predetermined area, the data analysis system receives and analyzes the heat generation data to determine whether the amount of heat generated in the predetermined area exceeds current settings of the tunable temperature controller, and in response to the amount of heat generated exceeding the current settings of the tunable temperature controller, updating the current settings to provide an amount of cooling that compensates for the amount of heat generated in the predetermined area in association with the electronic equipment.

Abstract: A device that includes a node engine configured to emulate a first node, a second node, and a third node. The first node is configured to receive a first correlithm object, fetch a second correlithm object based on the first correlithm object, and output the second correlithm object to the second node and the third node. Each correlithm object is a point in an n-dimensional space represented by a binary string. The second node is configured to receive the second correlithm object, fetch a third correlithm object based on the second correlithm object, and output the third correlithm object to the third node. The third node is configured to receive the second correlithm object, receive the third correlithm object, fetch a fourth correlithm object based on the second correlithm object and the third correlithm object, and output the fourth correlithm object.

Abstract: An on-demand transport system can manage an on-demand transportation service for a given region by matching requesting users with drivers and the AVs, where the AVs utilize localization maps and live sensor data to autonomously operate throughout the given region. The transport system can identify a local anomaly within the given region that affects AV performance. The transport system can transmit a routing invitation a driver to provide feedback corresponding to the local anomaly. Based on feedback data received from the driver, the transport system can transmit an update to AVs intersecting the local anomaly to enable the intersecting AVs to resolve the local anomaly.

Abstract: A method and system for image processing is provided. The technique includes acquiring data related to the processing, performing a pre-processing of the data, performing a segmentation of a subject, performing a post-processing of the result of the segmentation and managing storage of the data. The post-process includes a calibrating and a rendering of the data.

Abstract: A modular utility room a mobile enclosure is positionable adjacent to an information technology (IT) enclosure of a modular data center (MDC). One of a first and second equipment cabinet is selected for installing in the mobile enclosure. Each equipment cabinet contains functional equipment having a lower electrical coupling component that are connectable to substrate-installed conduits below the mobile enclosure and an upper electrical coupling component that connect to and support rack-based information handling systems (IHSs) in the IT enclosure. An interface panel is received in an interface panel opening in the mobile enclosure correspond to the selected equipment cabinet and will not receive the other equipment cabinet. The interface panel positions the lower electrical coupling over a conduit opening in alignment with a defined location relative to the mobile enclosure for alignment with the substrate-installed conduits at a deployed location of the mobile utility room.

Abstract: An advertising campaign received by an online system has a specified budget, a specified duration, and includes multiple advertisement requests that each include advertisements for presentation to users of the online system. An ad request included in the advertising campaign is associated with a frequency limit specifying a maximum number of times an advertisement from the ad request is shown to a user during the specified duration. When selecting advertisements for presentation to a user, the online system determines an adjustment value for the ad request's bid amount based on a number of times the advertisement from the ad request has been presented to the user, an amount of the duration that has lapsed, and the frequency limit associated with the ad request. The online system modifies the ad request's bid amount by the adjustment value and uses the modified bid amount when selecting advertisements for presentation to the user.

Abstract: A feedback value of a user is received, the feedback value corresponding to a content of a current post on a social media service. The content is analyzed to determine a context of the current post. A range of feedback values is computed corresponding to the context. When the feedback value is outside the range of feedback values, a set of contents is computed, the set of contents spanning a period prior to a time of the current post, and the set of contents corresponding to the context. The set of contents omits another post that is within the period and is related to the current post. A recap timeline is constructed using the set of contents and rendered relative to the current post.

Abstract: An operation monitoring server includes a storage apparatus configured to store instructions, and a processor configured to operate based on the instructions. The processor is configured to acquire state information on at least one of a monitoring target device or a base location in which the monitoring target device is placed. The processor is configured to select, based on the acquired state information, one of a plurality of service request processing apparatus each having a function of processing a service request received from the monitoring target device.

Abstract: A system and method for providing a search experience in which users are protected from exposure to inappropriate offensive or sensitive content is described. A search system may classify a search query and candidate search results obtained in response to the search query. Based on the classification of the search query and search results, the candidate search results may be modified to generate a set of search results presented to a user such that the presented search results do not include inappropriate sensitive or offensive content.

Abstract: A method for managing a cache memory network is disclosed. The method comprises estimating the popularity of a piece of data content according to an extreme learning machine, and managing the cache memory network according to the estimated popularity of the piece of data content. Estimation may be performed upon a request for the piece of data content, and may be accordingly cashed into an appropriate cache memory device upon estimation of its popularity.

Abstract: A method for developing machine operation classifiers for a machine is disclosed. The method includes receiving training data associated with the machine from one or more on-board engineering channels associated with the machine and determining one or more training features based on the training data values. The method also includes determining one or more training labels associated with the one or more training features and building a predictive model for determining machine operation classifiers using a computer. Building the predictive model may include feeding the one or more training features and the one or more training labels associated with the one or more training features to a machine learning algorithm and determining a predictive model from the machine learning algorithm. The predictive model may be used for receiving new data associated with the machine and determining a predicted label based on the new data.

Abstract: Systems and methods that utilize an image processing model to analyze features of an image to identify individuals within a set of images are disclosed. Feature vectors for the images are extracted and clustered based on a distance between the feature vectors in a feature space. The clusters are used to identify individuals within the images with the aid of human annotators. The identified individuals and images can be used to retrain the image processing model.

Abstract: A device that includes a node engine configured to emulate a first node, a second node, and a third node. The first node is configured to receive a first correlithm object, fetch a second correlithm object based on the first correlithm object, and output the second correlithm object to the second node and the third node. Each correlithm object is a point in an n-dimensional space represented by a binary string. The second node is configured to receive the second correlithm object, fetch a third correlithm object based on the second correlithm object, and output the third correlithm object to the third node. The third node is configured to receive the second correlithm object, receive the third correlithm object, fetch a fourth correlithm object based on the second correlithm object and the third correlithm object, and output the fourth correlithm object.

Abstract: One embodiment of the invention provides a system for mapping a neural network onto a neurosynaptic substrate. The system comprises a metadata analysis unit for analyzing metadata information associated with one or more portions of an adjacency matrix representation of the neural network, and a mapping unit for mapping the one or more portions of the matrix representation onto the neurosynaptic substrate based on the metadata information.

Abstract: Techniques to identify application foreground/background state based on network traffic are described. In one embodiment, an apparatus may comprise a traffic monitoring component and a traffic analysis component. The traffic monitoring component may receive monitored network traffic over a monitored time period. The traffic analysis component may receive a foreground activity profile, the foreground activity profile comprising one or more signals for distinguishing between foreground activity of one or more profiled applications and background activity of the one or more profiled applications; determine one or more active foreground applications in the monitored network traffic based on the foreground activity profile, wherein determining the one or more active foreground applications comprises determining that at least a portion of the monitored network traffic comprises background activity; and log that the one or more active foreground application were active during the monitored time period.

Abstract: An Expandable Modular Information Technology (EMIT) Facility (EMITF) includes: an EMIT Building Infrastructure (EMITBI) having an initial enclosure that includes: a base pad providing a fixed usable space area on which one or more modular IT components are placed; and a plurality of exterior walls, including at least one removable expansion wall (REW) that enables expansion of the usable space area of the EMITBI; a plurality of modular IT components placed within the EMITBI on the usable space, and which dissipate heat; and a power subsystem electrically coupled to the EMITBI via a power distribution configuration that includes a first power generating component coupled via a power distribution scheme to the power consuming IT components of the EMITBI, where the power subsystem is pre-configured to support additional requirements for power via a power capacity on demand structure, including modular expandability of the power generating components, while maintaining resiliency.

Abstract: Building models and predicting operational outcomes of a drilling operation. At least some of the illustrative embodiments are methods including: gathering sensor data regarding offset wells and context data regarding the offset wells, and placing the sensor data and context data into a data store; creating a reduced data set by identifying a correlation between data in the data store and an operational outcome in a drilling operation; creating a model based on the reduced data set; and predicting the operational outcome based on the model.

Abstract: A source deconvolutional network is adaptively trained to perform semantic segmentation. Image data is then input to the source deconvolutional network and outputs of the S-Net are measured. The same image data and the measured outputs of the source deconvolutional network are then used to train a target deconvolutional network. The target deconvolutional network is defined by a substantially fewer numerical parameters than the source deconvolutional network.

Abstract: The disclosed embodiments relate to a system that selectively propagates information through a neuromorphic circuit comprising a set of interconnected neurons. During operation, a neuron in the set of neurons receives information-carrying current pulses from one or more upstream information-carrying neurons, wherein the information-carrying current pulses are insufficient to cause the neuron to generate output current pulses. The neuron also receives selectively generated gating current pulses from one or more gating neurons, wherein the gating current pulses cause a neural voltage of the neuron to approach a firing threshold. In this way, concurrently received information-carrying current pulses combine with the gating current pulses to cause the neural voltage to exceed the firing threshold, which causes the neuron to generate output current pulses that propagate to downstream neurons.

Abstract: For deduplicated data processing hierarchical rate control in a data deduplication system in a computing storage environment, priorities in a hierarchal structure are propagated by defining a relationship between each of a plurality of maintenance tasks and using a resource distribution function for managing and dividing resources between the plurality of maintenance tasks based on a priority for each maintenance task.

Abstract: A neural system comprises multiple neurons interconnected via synapse devices. Each neuron integrates input signals arriving on its dendrite, generates a spike in response to the integrated input signals exceeding a threshold, and sends the spike to the interconnected neurons via its axon. The system further includes multiple noruens, each noruen is interconnected via the interconnect network with those neurons that the noruen's corresponding neuron sends its axon to. Each noruen integrates input spikes from connected spiking neurons and generates a spike in response to the integrated input spikes exceeding a threshold. There can be one noruen for every corresponding neuron. For a first neuron connected via its axon via a synapse to dendrite of a second neuron, a noruen corresponding to the second neuron is connected via its axon through the same synapse to dendrite of the noruen corresponding to the first neuron.