Abstract: A computer system configures processing elements within a distributed computing system. A processing element within a distributed computing environment is determined to be affected by a software update, wherein each processing element of the distributed computing system includes a plurality of components and the software update modifies the components of the processing elements. The determined processing element is split into a plurality of processing elements based on a set of factors. The plurality of components of the processing element are assigned among the plurality of processing elements based on components affected by the software update. Embodiments of the present invention further include a method and program product for configuring processing elements within a distributed computing system in substantially the same manner described above.

Abstract: An object classifier using a set of object definitions arranged in an object definition hierarchy including at least a first group of coarse-level object definitions and a second group of finer-level object definitions. The object classifier is arranged to configure a first object classification cycle and a second, subsequent, object classification cycle by selectively executing a first subset of object definitions from the categorization data in the first object classification cycle; and selectively executing a second, different, subset in the second object classification cycle.

Abstract: The present disclosure relates to a system, a method, and a non-transitory computer readable storage medium for deep packet inspection scanning at an application layer of a computer. A method of the presently claimed invention may scan pieces of data received out of order without reassembly at an application layer from a first input state generating one or more output states for each piece of data. The method may then identify that the first input state includes one or more characters that are associated with malicious content. The method may then identify that the data set may include malicious content when the first input state combined with one or more output states matches a known piece of malicious content.

Abstract: A data analytical engine receives packets from a number of different network interface devices. The data is a replica of part or all of transmit or receive packets processed in the network interface device. A learning algorithm is applied to data from said different network interface devices and it is determined if an alert is to be generated.

Abstract: A system evaluates a plurality of faults in an operation of a machine at a set of future instances of time. The system uses a neural network including a first subnetwork sequentially connected with a sequence of second subnetworks for each of the future instance of time such that an output of one subnetwork is an input to a subsequent subnetwork. The first subnetwork accepts the current time-series data and the current setpoints of operation of the machine. Each of the second subnetworks accepts the output of a preceding subnetwork, an internal state of the preceding subnetwork, and a future setpoint for a corresponding future instance of time. Each of the second subnetworks outputs an individual prediction of each fault of a plurality of faults at the corresponding future instance of time.

Abstract: An embodiment of the invention may include a method, computer program product, and system for generating ground truth data for a plurality of cognitive capabilities within an overall cognitive system. The embodiment may include configuring multiple sets of training data. Each set of training data corresponds to a separate cognitive capability. The embodiment may include displaying a set of ground truth curation activities via a user interface. The embodiment may include determining the ground truth curation activities performed for a first type of data for a first duration. The first type of data is selected from the single set of grouped training data. The embodiment may include determining whether the first duration has exceeded a pre-determined threshold. The embodiment may include switching the curation activities to a second type of data. The second type of data is selected from the single set of grouped data.

Abstract: Systems and methods are described for matching a corrupted database record with a record of a validated database. The system receives a corrupted record from a first database. The corrupted record is vectorized to create an input data vector. A denoised data vector is generated by applying a denoising autoencoder to the input data vector, where the denoising autoencoder is specific to the first database. The system compares the denoised data vector with each of a plurality of validated data vectors generated based on records of the validated database to determine that a first denoised data vector matches a matching vector. In response, the system trains the denoising autoencoder using a data pair that includes the input data vector and the matching vector. The system also outputs the validated record that was used to generate the first matching vector.

Abstract: This disclosure is directed to methods and systems for managing client-side services. An intermediary may receive from an agent incorporated into a web application or client-side application and executing on the client device, a request for a configuration to manage loading of services from the at least one server onto the web application or client-side application. The intermediary can determine an expected temporal pattern of use for at least some of the services at the client device. The intermediary can generate the configuration. The configuration can specify for each of the services, an expected amount of bandwidth to be utilized for loading the corresponding service from the at least one server onto the web application or client-side application, and a loading priority according to the expected temporal pattern of use. The intermediary can send the configuration to the agent.

Abstract: Minimizing the latency of on-device detection of malicious executable files, without sacrificing accuracy, by applying a machine learning model to an executable file in quantized steps. Allowing a threshold confidence level to be set to different values enables controlling the tradeoff between accuracy and latency in generating a confidence level indicative of whether the executable file includes malware.

Abstract: A method, a computer system, and a computer program product for selecting electronic documents based on an expertness level for a target keyword. The method includes identifying a Lowest Common Ancestor (LCA) for each pair of categories to which documents in a document set belong based on a graph having edges which define relationships between nodes representing respective ones of the categories, the document set being associated with a target keyword, obtaining a degree of dispersion of the categories based on respective distances between the categories of each pair and the corresponding LCA, determining an expertness level for the target keyword based on the degree of dispersion, and in response to receiving a search query including the target keyword, selecting one or more documents from a database based on the expertness level for the target keyword.

Abstract: A computer-implemented method includes obtaining a training data set including a plurality of training examples. The method includes generating, for each training example, multiple feature vectors corresponding, respectively, to multiple feature types. The method includes applying weighting factors to feature vectors corresponding to a subset of the feature types. The weighting factors are determined based on one or more of: a number of training examples, a number of classes associated with the training data set, an average number of training examples per class, a language of the training data set, a vocabulary size of the training data set, or a commonality of the vocabulary with a public corpus. The method includes concatenating the feature vectors of a particular training example to form an input vector and providing the input vector as training data to a machine-learning intent classification model to train the model to determine intent based on text input.

Abstract: A method, computer system, and a computer program product for assessing anonymity of a dataset is provided. The present invention may include receiving an original dataset and an anonymized dataset. The present invention may also include preparing a testing dataset and a training dataset for a machine learning algorithm based on the received original dataset and anonymized dataset. The present invention may then include training a machine learning model based on the prepared training dataset. The present invention may further include generating an evaluation score based on the trained machine learning model and the prepared testing dataset. The present invention may also include presenting the generated evaluation score to a user.

Abstract: A cognitive meeting assistant facility is provided for assisting a moderator in an online meeting. The cognitive meeting assistant facility automatically categorizes chat questions of participants of the online meeting based on content into multiple categories of chat questions. One or more categories of chat questions of the multiple categories of chat questions include multiple questions of different participants of the online meeting. Further, the cognitive meeting assistant facility deploys the multiple categories of chat questions for display to the participants of the online meeting. The deploying includes, for each category of the at least one category of chat questions, providing a category-applicable answer for display to the participants of the online meeting in association with that category of chat questions.

Abstract: A computer system configures processing elements within a distributed computing system. A processing element within a distributed computing environment is determined to be affected by a software update, wherein each processing element of the distributed computing system includes a plurality of components and the software update modifies the components of the processing elements. The determined processing element is split into a plurality of processing elements based on a set of factors. The plurality of components of the processing element are assigned among the plurality of processing elements based on components affected by the software update. Embodiments of the present invention further include a method and program product for configuring processing elements within a distributed computing system in substantially the same manner described above.

Abstract: Methods, systems, and techniques for performing object detection using a convolutional neural network (CNN) involve obtaining an image and then processing the image using the CNN to generate a first feature pyramid and a second feature pyramid from the first pyramid. The second pyramid includes an enhanced feature map, which is generated by combining an upsampled feature map and a feature map of the first feature pyramid that has a corresponding or lower resolution of a resolution of the enhanced feature map. The upsampled feature map is generated by upsampling a feature map of the second feature pyramid that is at a shallower position in the CNN than the enhanced feature map. The enhanced feature map is split into channel feature maps of different resolutions, with each of the channel feature maps corresponding to channels of the enhanced feature map. Object detection is performed on the channel feature maps.

Abstract: Described embodiments provide systems and methods for inferring a network type and network conditions. The system includes a packet capturing engine configured to capture a plurality of network packets from a plurality of TCP network connections. The system includes a packet analyzer configured to analyze the plurality of network packets to generate a plurality of metrics. The system includes a network classifier configured to infer network types of the plurality of TCP connections based on the plurality of metrics and at least one classification model. The system also includes a conditions ranking engine configured to estimate a level of network congestion for each TCP connection based on the plurality of metrics and the network types.

Abstract: A color predictor is provided to predict the color of a food item given its formula comprising the ingredients and its quantities. The color predictor may utilize machine learning algorithms and a set of recipe data to train the color predictor. The color predictor can also be used by a color recommender to recommend changes in the given formula to achieve a target color.

Abstract: In one embodiment, a method includes receiving, from a first sensor on a robot, first sensor data indicative of an environment of the robot. The method also includes identifying, based on the first sensor data, an object of an object type in the environment of the robot, where the object type is associated with a classifier that takes sensor data from a predetermined pose relative to the object as input. The method further includes causing the robot to position a second sensor on the robot at the predetermined pose relative to the object. The method additionally includes receiving, from the second sensor, second sensor data indicative of the object while the second sensor is positioned at the predetermined pose relative to the object. The method further includes determining, by inputting the second sensor data into the classifier, a property of the object.

Abstract: Discloses are systems, methods and computer readable mediums for automated verifications of potential vulnerabilities of one or more sites or code utilizing one or more neural networks. The systems, methods and computer readable mediums can transmit one or more scan operations to one or more sites, receive one or more responses to the one or more scan operations, tokenize the one or more responses, transmit to one or more neural networks the one or more tokenized responses, receive from the one or more neural networks verification of the one or more tokenized responses, and determine one or more confidences of the one or more verified responses.

Abstract: A method and system for determining the hierarchical purchase decision process of a shopper in front of a product category in a retail store. Shopping consideration and the decision path the shopper can be obtained by combining behavior data with the category layout and transaction data. The hierarchical decision process can be determined based on observed actual in-store purchase behavior using a set of video cameras and processor implemented instructions for extracting the sequence and timing of the shopper's decision process. The hierarchical decision process, obtained by the clustering of shopper behavior data over multiple shopping trips, can not only identify the sequence of the decision, but can also quantify the volume of shopping trips at each level in the decision process and the amount of time spent by the shopper for making each decision. The decisions of the shopper for each product group can be captured down to the product attribute level.

Abstract: An online system enforces policies to content items that are distributed on its platform and blocks content items that violate one or more of those policies. To identify content items that are slightly varied from each other, the online system generates an embedding for each of the known content items that have already been determined to be noncompliant with one or more policies. The online system then groups the known noncompliant content items that are clustered together in the embedding space. The texts of the group of known noncompliant content items are converted to finite state automata and are merged to generate a common automaton. The common automaton is used to generate a common regular expression that is used to screen new content items. When a new content item matches the textual pattern defined by the common regular expression, the system may block the new content item.

Abstract: Prediction, detection and mitigation of network or device issues in a communication system are facilitated. An embodiment can comprise: determining whether an identified problem of a device has an associated defined solution stored in a repository of information; transmitting solution information representative of the associated defined solution to the device for application of the solution information to the device in a manner determined to have less than a defined amount of impact on the device and in accordance with defined security protocols of the device; and assessing a performance of the device after application of the solution information to the device to determine whether the solution information solved the identified problem. In some embodiments, solution detection can be performed such that based on a determination that the identified problem has been removed, the change that caused the identified problem can be determined.

Abstract: Disclosed are a method, a device, a system and/or a manufacture of application of a spreadsheet formula algorithm against a dataset such as a large external data source. In one embodiment, a scalable method of analyzing data includes generating a prototype data through importing a data entry from the dataset and mapping to cells of a spreadsheet file that may be accessible as a software-as-a-service. A data model for analyzing the dataset is defined through a spreadsheet algorithm comprising spreadsheet formulas outputting a dependent variable. The spreadsheet formulas, with one or more independent variables as inputs, are stored in a syntax format permitting independent calculation of the dependent variable. An extrapolated algorithm expressed in a programming language that may include SQL is generated from the formula algorithm and applied against the dataset utilizing parallel processing to generate a value for the dependent variable of each data entry of the dataset.

Abstract: Methods and systems for performing a sequence of machine learning tasks. One system includes a sequence of deep neural networks (DNNs), including: a first DNN corresponding to a first machine learning task, wherein the first DNN comprises a first plurality of indexed layers, and each layer in the first plurality of indexed layers is configured to receive a respective layer input and process the layer input to generate a respective layer output; and one or more subsequent DNNs corresponding to one or more respective machine learning tasks, wherein each subsequent DNN comprises a respective plurality of indexed layers, and each layer in a respective plurality of indexed layers with index greater than one receives input from a preceding layer of the respective subsequent DNN, and one or more preceding layers of respective preceding DNNs, wherein a preceding layer is a layer whose index is one less than the current index.

Abstract: A system and method to simulate the impact of leadership activity for individuals in leader and/or manager roles to be able to model their organization as a system and to simulate a plurality of actions that might be taken and their impact on the non-linear dynamics of the organization, its functions, capabilities, processes and outcomes. The system includes an Organization State component, a Leadership Activities component, an Environment State component, and a Simulation Module component. The Simulation Module component takes in initial conditions defined for the Organization State, Leadership Activities and the Environment State, iterates the functions and process of the organization through time, simulates the interactions of the various variables described in the Organization State, the Leadership Activities and the Environment State and the interactive effects among them and provides outputs of a plurality of measures for each time step.

Abstract: Example embodiments for classification are described. In an example embodiment, a request including a text term to be classified is received from a source system via a communication network at a computer system. A rule associated with the text term is accessed, in which the rule indicates at least one classifier of a plurality of classifiers installed at the computer system. Classification information for the text term is generated at the computer system using the at least one classifier indicated by the rule. The generated classification information includes a classification selected from a taxonomy by the at least one classifier. The generated classification information is transmitted via the communication network to the source system.

Abstract: An approach is described for adjusting prominence of a participant profile in a social networking interface. An associated method may include receiving an activity stream update of the participant and calculating a relevancy score based on content in the activity stream update. The method further may include adjusting a visibility level of the participant profile in the social networking interface based upon the calculated relevancy score. Adjusting the visibility level may include increasing the visibility level of the participant profile upon determining that the calculated relevancy score is greater than or equal to a first predefined threshold value. Adjusting the visibility level further may include decreasing the visibility level of the participant profile upon determining that the calculated relevancy score is less than a second predefined threshold value.

Abstract: A pirated video identification system for network operators to detect and identify both encrypted and unencrypted Internet protocol (IP) pirated video. The system uses ground truth data gathered from various video player clients as input to train a machine learning that then is used to analyze both real-time and historical IP flow data from a network to identify the IP addresses of both the source and destination of pirate video traffic on the network. They system can be used to aid in accurately measuring the volume and scale of pirated video traffic on a network as well as aiding in taking countermeasures against both the distributors and consumers of the pirated video.

Abstract: Techniques related to key phrase detection for applications such as wake on voice are discussed. Such techniques may include updating a start state based rejection model and a key phrase model based on scores of sub-phonetic units from an acoustic model to generate a rejection likelihood score and a key phrase likelihood score and determining whether received audio input is associated with a predetermined key phrase based on the rejection likelihood score and the key phrase likelihood score.

Abstract: An approach is described for adjusting prominence of a participant profile in a social networking interface. An associated method may include receiving an activity stream update of the participant and calculating a relevancy score based on content in the activity stream update. The method further may include adjusting a visibility level of the participant profile in the social networking interface based upon the calculated relevancy score. Adjusting the visibility level may include increasing the visibility level of the participant profile upon determining that the calculated relevancy score is greater than or equal to a first predefined threshold value. Adjusting the visibility level further may include decreasing the visibility level of the participant profile upon determining that the calculated relevancy score is less than a second predefined threshold value.

Abstract: Some embodiments provide a novel method for monitoring network requests from a machine. The method captures the network request at various layers of a protocol stack. At a first layer of a protocol stack, the method tags a packet related to the network request with a tag value, maps the tag value to a set of tuples associated with the packet, and sends a first set of data related to the packet to a security engine. At a second layer of the network stack, the method determines whether the packet has been modified through the network stack, and sends an updated second set of data to the security engine when the packet has been modified.

Abstract: A method of creating characteristic peak profiles of mass spectra and identification model for analyzing and identifying microorganisms are provided. MALDI-TOF MS data of microorganisms having the same feature are gathered. Discretization of the data is performed. Density-based clustering is used to find m/z values of spectral peaks with high probability of occurrence from the discretized data. A characteristic MS peak profile is created for every specific feature of microorganisms. Every such a characteristic profile forms a feature template. The mass spectrum of each known isolate is matched against all the feature templates and a number of matched vectors are obtained. The matched vectors are then concatenated into a single “integrated vector.” Then, a machine learning method and the integrated vectors generated from all known isolates are used to create a classification model for microorganism identification.

Abstract: Various embodiments described herein relate to methods and apparatus for robust classification. Many real-world datasets suffer from missing or incomplete data. By assigning weights to certain features of a dataset based on which feature(s) are missing or incomplete, embodiments of the prevention can provide robustness and resilience to missing data.

Abstract: Systems, methods, and computer-readable media are disclosed for determining recommended sizes of products using a machine learning model.

Abstract: There is provided a system for computing connectivity of medical imaging network nodes, comprising: at least one hardware processor executing a code for: monitoring packets transmitted over a network connecting a plurality of network nodes, analyzing the monitored packets to identify packets associated with a Digital Imaging and Communications in Medicine (DICOM) protocol, analyzing the packets associated with the DICOM protocol to identify medical imaging network nodes implementing the DICOM protocol, designating a respective medical imaging type for each node of the medical imaging network nodes, and computing a data structure storing connectivity of the medical imaging network nodes each designated with the respective medical imaging type.

Abstract: A cognitive analysis computer device is programmed to a) receive a set of streaming input data from one or more sensors; b) perform spatial abstraction on the set of streaming input data to divide the set of streaming input data into a plurality of input pathways; c) route the set of streaming input data from a plurality of input pathways to a plurality of output pathways based on a context mapping; d) for each output pathway, perform temporal abstraction on the received input values to generate an output value for the corresponding output pathway by accumulating output values from a plurality of sets of streaming input data until at least one output value of the plurality of output values on the plurality of output pathways exceeds a predetermined threshold; and e) compare the plurality of output values to one or more stored sets of data to determine a match.

Abstract: A search system includes a reception unit 601 for receiving specification of a specific sentence in an article, a search unit 602 using a first sentence being the specified sentence as a search query to search an article including a sentence having a meaning similar to the connotations of the first sentence, and a display unit 603 for displaying information of the searched article as a related article to the first sentence, and when there are a plurality of related articles, the display unit 603 preferentially displays the related articles in descending order of difference relative to a first article being the article including the first sentence.

Abstract: A system and method of automatically assigning a priority rank to messages. The system and method accesses a message data store and assigns a priority rank to each message. The priority rank is selected from a priority rank scale by, for each message, parsing the message for features present in the message and calculating a predicted intensity score for the message using a user-specific classifier. The classifier is trained from user training data which includes prior user messages on which a machine learning algorithm operates. The training data is labeled by scores calculated based on the actual activates performed by the user to each message. The priority rank of each message can be used to improve message processing in message processing systems.

Abstract: A method is provided for prediction of a physical divider on a road. The method comprises retrieving vehicular sensor data, map data, or a combination thereof captured over a predefined period of time for at least one segment of the road. The method further comprises aggregating the respective vehicular sensor data, map data, or the combination thereof for the at least one segment of the road to generate one or more aggregated values. The method further comprises generating output data corresponding to presence of the physical divider in the at least one segment of the road by a machine learning model based on the one or more aggregated values as an input to the machine learning model.

Abstract: An event detection system is configured to access a repository that contains a collection of media content. The media content may for example include images, videos, audio clips, and the like, wherein the media content comprises features that include: tags (e.g., hashtags or other similar mechanisms to label and sort content); captions that comprises one or more words or phrases; continuous numerical values; geolocation data (e.g., geo-hash, check-in data, coordinates); as well as temporal data (e.g., timestamps).

Abstract: An information processing apparatus accepts information corresponding to a text. The information processing apparatus refers to a storage unit that stores therein co-occurrence information on other texts with respect to the text and information corresponding to the other texts by associating both the information with the text. The information processing apparatus specifies, from among the pieces of information corresponding to the other texts, the text associated with the information corresponding to the other texts that is associated with the co-occurrence information that meets the standard.

Abstract: According to an embodiment, an information processing apparatus includes a memory and processing circuitry. The processing circuitry configured to acquire a plurality of input images captured at a specific place. The processing circuitry configured to adapt an estimation model used for detecting a target object included in images to the specific place based on the plurality of input images. The processing circuitry configured to output a result of determination of an adaptation state for the specific place in the estimation model.

Abstract: Described herein are technologies pertaining to determining which search engine results page (SERP), from a plurality of SERPs, is preferable to a user. A query is received, and multiple SERPS are retrieved based upon the query, wherein the multiple SERPs are generated independently from one another. Values of features of the query and the multiple SERPs are obtained, and a determination as to which of the SERPs from the multiple SERPs is preferable to the user is made based upon the values of the features. The SERP determined to be preferable over other SERPs in the multiple SERPs is presented to the user.

Abstract: The present invention provides an event-driven universal neural network circuit. The circuit comprises a plurality of neural modules. Each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module. An interconnection network comprising a plurality of digital synapses interconnects the neural modules. Each synapse interconnects a first neural module to a second neural module by interconnecting a neuron in the first neural module to a corresponding neuron in the second neural module. Corresponding neurons in the first neural module and the second neural module communicate via the synapses. Each synapse comprises a learning rule associating a neuron in the first neural module with a corresponding neuron in the second neural module. A control module generates signals which define a set of time steps for event-driven operation of the neurons and event communication via the interconnection network.

Abstract: A neural network is structured with a plurality of levels of nodes. Each level has a level-specific stabilization parameter that adjusts a learning rate, at a corresponding level, during training. The stabilization parameter has a value that varies inversely relative to a change in an objective training function during back-propagation of the error through the level.

Abstract: A method for predicting a concentration of an in vivo analyte includes obtaining a plurality of in vivo spectra of the in vivo analyte, determining a learning section for a concentration predicting algorithm for the analyte based on an unchanged section, during which the concentration of the analyte is not substantially changed, and a plurality of the in vivo spectra, and predicting the concentration of the in vivo analyte by using the concentration predicting algorithm based on a learned result of the learning section and an intrinsic spectrum of the in vivo analyte.

Abstract: A classifier 81 classifies target data into a cluster on the basis of a mixture model defined using two different types of variables that indicate features of the target data. In this classification, the classifier 81 classifies the target data into a cluster on the basis of a mixture model in which a mixing ratio of the mixture model is represented by a function of a first variable and in which the element distribution of the clusters into which the target data is classified is represented by a function of a second variable.

Abstract: The present disclosure provides systems and methods that perform stochastic generative hashing. According to one example aspect, a machine-learned hashing model that generates a binary hash for an input can be trained in conjunction with a machine-learned generative model that reconstructs the input from the binary hash. The present disclosure provides a novel generative approach to learn hash functions through Minimum Description Length principle such that the learned hash codes maximally compress the dataset. According to another example aspect, the present disclosure provides an efficient learning algorithm based on the stochastic distributional gradient, which avoids the notorious difficulty caused by binary output constraints, to jointly optimize the parameters of the hashing model and the associated generative model.

Abstract: Systems and methods for automated segmentation of anatomical structures, such as the human heart. The systems and methods employ convolutional neural networks (CNNs) to autonomously segment various parts of an anatomical structure represented by image data, such as 3D MRI data. The convolutional neural network utilizes two paths, a contracting path which includes convolution/pooling layers, and an expanding path which includes upsampling/convolution layers. The loss function used to validate the CNN model may specifically account for missing data, which allows for use of a larger training set. The CNN model may utilize multi-dimensional kernels (e.g., 2D, 3D, 4D, 6D), and may include various channels which encode spatial data, time data, flow data, etc. The systems and methods of the present disclosure also utilize CNNs to provide automated detection and display of landmarks in images of anatomical structures.

Abstract: A network interface device has in input. The input receives packets in accordance with a protocol and has at least one protocol header. The network interface has hardware which applies an artificial intelligence process to at least one of the protocol headers. This is used to provide an output which may, for example, indicate a risk associated with a packet.