Abstract: An exemplary discovery platform includes machine-learning techniques for using medical imaging data to study a phenotype of interest, such as complex diseases with weak or unknown genetic drivers. An exemplary method of identifying a patient subgroup of interest, comprises inputting a plurality of medical images obtained from a group of clinical subjects into a trained unsupervised machine-learning model to obtain a plurality of embeddings in a latent space, clustering the plurality of embeddings to generate one or more clusters of embeddings, identifying one or more patient subgroups corresponding to the one or more clusters of embeddings, and associating each patient subgroup of the one or more patient subgroups with a covariant to identify the patient subgroup of interest.

Abstract: A firewall monitors network activity and stores information about that network activity in a network activity log. The network activity is analyzed to identify a potential threat. The potential threat is further analyzed to identify other potential threats that are related to the potential threat, and are likely to pose a future risk to a protected network. A block list is updated to include the potential threat and the other potential threats to protect the protected network from the potential threat and the other potential threats.

Abstract: An apparatus for generating a generalized linear model structure definition by generating a gradient boosted tree model and separating each decision tree into a plurality of indicator variables upon which a dependent variable of the generalized linear model depends.

Abstract: An inference system trains and performs inference using a sparse neural network. The sparse neural network may include one or more layers, and each layer may be associated with a set of sparse weights that represent sparse connections between nodes of a layer and nodes of a previous layer. A layer output may be generated by applying the set of sparse weights associated with the layer to the layer output of a previous layer. Moreover, the one or more layers of the sparse neural network may generate sparse layer outputs. By using sparse representations of weights and layer outputs, robustness and stability of the neural network can be significantly improved, while maintaining competitive accuracy.

Abstract: A method and system for vetting items being shipped across national boundaries using a new technology enabling an automated system is provided. The automated system screens items for shipping through customs and validating the items for shipment according to customs rules and regulations. The system identifies and applies the appropriate rules for customs and other responsible agencies pertaining to the eligibility of any item being imported into a particular country. The present invention utilizes an unconventional combination of image recognition technology, machine learning algorithms, and rule engine algorithms to categorize, identify, and apply the appropriate rules to each and every item being considered for importation to another country.

Abstract: Techniques and systems for a security service system configured with a sensor component including a machine learning (ML) malware classifier to perform behavioral detection on host devices. The security service system may deploy a sensor component to monitor behavioral events on a host device. The sensor component may generate events data corresponding to monitored operations targeted by malware. The system may map individual events from events data onto a behavioral activity pattern and generate process trees. The system may extract behavioral artifacts to build a feature vector used for malware classification and generate a machine learning (ML) malware classifier. The sensor component may use the ML malware classifier to perform asynchronous behavioral detection on a host device and process system events for malware detection.

Abstract: A speech recognition platform configured to receive an audio signal that includes speech from a user and perform automatic speech recognition (ASR) on the audio signal to identify ASR results. The platform may identify: (i) a domain of a voice command within the speech based on the ASR results and based on context information associated with the speech or the user, and (ii) an intent of the voice command. In response to identifying the intent, the platform may perform a corresponding action, such as streaming audio to the device, setting a reminder for the user, purchasing an item on behalf of the user, making a reservation for the user or launching an application for the user. The speech recognition platform, in combination with the device, may therefore facilitate efficient interactions between the user and a voice-controlled device.

Abstract: Methods, apparatus, systems and articles of manufacture are disclosed that optimize workflows. An example apparatus includes an intent determiner to determine an objective of a user input, the objective indicating a task to be executed in an infrastructure, a configuration composer to compose a plurality of workflows based on the determined objective, a model executor to execute a machine learning model to create a confidence score relating to the plurality of workflows, and a workflow selector to select at least one of the plurality of workflows for execution in the infrastructure, the selection of the at least one of the plurality of workflows based on the confidence score.

Abstract: An imaging device having a function of processing an image is provided. The imaging device has an additional function such as image processing, can hold analog data obtained by an image capturing operation in a pixel, and can extract data obtained by multiplying the analog data by a predetermined weight coefficient. Difference data between adjacent light-receiving devices can be obtained in a pixel, and data on luminance gradient can be obtained. When the data is taken in a neural network or the like, inference of distance data or the like can be performed. Since enormous volume of image data in the state of analog data can be held in pixels, processing can be performed efficiently.

Abstract: The present disclosure is directed toward systems, methods, and non-transitory computer readable media for generating and searching a hybrid search index. In some embodiments, the disclosed systems generate a hybrid search index that comprises one or more content items stored at a content management system or at external network locations linked to the content management system via software connectors along with world state data associated with the one or more content items. The disclosed systems can generate a search result from the hybrid search index in response to receiving a search query of the hybrid search index. In some cases, the disclosed systems can rank one or more content items included in the search result based on observation layer data of the one or more content items.

Abstract: A system can receive a first set of data. The first set of data can include information indicating a first set of user sessions and for each of the first set of user sessions having an associated summary and a corresponding agent indicated intent. The system can also, based on the first set of data, determine a set of utterances and for each of the set of utterances a corresponding set of intents. Additionally, the system can receive a second set of data. The second set of data including information indicating a second set of user sessions and for each of the second set of user sessions having an associated determined utterance and corresponding interaction of a user. Moreover, the system can validate a corresponding intent of one or more utterances of the set of utterances, based on the second set of data.

Abstract: A method and system of training an interpretable deep learning model includes receiving an input set of data, which may be complex. The input set of data is provided to deep learning model for feature extraction. In an exemplary embodiment, the deep learning model generates a disentangled latent space of features from the feature extraction. The features may comprise semantically meaningful data which is then provided to a low-complexity learning model. The low-complexity learning model generates output based on a specified task (for example, classification or regression). Being a low-complexity learning model provides confidence that the data output from the deep learning model is inherently interpretable.

Abstract: A method of jointly training of a transferable feature extractor network, an ordinal regressor network, and an order classifier network in an ordinal regression unsupervised domain adaption network by providing a source of labeled source images and unlabeled target images; outputting image representations from a transferable feature extractor network by performing a minimax optimization procedure on the source of labeled source images and unlabeled target images; training a domain discriminator network, using the image representations from the transferable feature extractor network, to distinguish between source images and target images; training an ordinal regressor network using a full set of source images from the transferable feature extractor network; and training an order classifier network using a full set of source images from said transferable feature extractor network.

Abstract: According to some embodiments, a method performed by a classification scanner comprises receiving an electronic message and determining a classification that applies to the electronic message. The classification is determined based on an express indication from a user. The method further comprises providing a machine learning trainer with the electronic message and an identification of the classification that applies to the electronic message. The machine learning trainer is adapted to determine a machine learning policy that associates attributes of the electronic message with the classification.

Abstract: The technology disclosed corrects inter-cluster intensity profile variation for improved base calling on a cluster-by-cluster basis. The technology disclosed accesses current intensity data and historic intensity data of a target cluster, where the current intensity data is for a current sequencing cycle and the historic intensity data is for one or more preceding sequencing cycles. A first accumulated intensity correction parameter is determined by accumulating distribution intensities measured for the target cluster at the current and preceding sequencing cycles. A second accumulated intensity correction parameter is determined by accumulating intensity errors measured for the target cluster at the current and preceding sequencing cycles. Based on the first and second accumulated intensity correction parameters, next intensity data for a next sequencing cycle is corrected to generate corrected next intensity data, which is used to base call the target cluster at the next sequencing cycle.

Abstract: A model may be trained on a training dataset, e.g., for medical image processing or medical signal processing tasks. Systems and computer-implemented methods are provided for associating a population descriptor with the trained model and using the population descriptor to determine whether records to which the model is to be applied, conform to the population descriptor. The population descriptor characterizes a distribution of the one or more characteristic features over the training dataset, with the characteristic features characterizing the training record and/or a model output provided when the trained model is applied to the training record. For instance, the model may be applied only to records conforming to population descriptor, or model outputs of applying the model to non-conforming records may be flagged as possibly untrustworthy.

Abstract: Disclosed is a few-shot gesture recognition method. The method comprises the following steps: customizing, by a user, gesture categories, and acquiring few samples for each gesture category; inputting the acquired samples into a trained few-shot learning model, extracting a feature vector corresponding to each sample, and synthesizing feature vectors belonging to the same gesture to obtain an average feature vector corresponding to each gesture as a prototype vector; acquiring a corresponding sample for a target gesture implemented by the user, and inputting the sample into the few-shot learning model to obtain a feature vector of the target gesture as a query vector; and calculating similarities between the query vector and prototype vectors of different gestures, and selecting a gesture category corresponding to the prototype vector with the highest similarity as a prediction category of the target gesture.

Abstract: A group masked autoencoder may be implemented for anomaly detection. An autoencoder network model may be trained without supervision and applied to output an estimated joint probability distribution of normality for a group of frames of time series data. The estimated joint probability distribution may be used to determine an anomaly score for the time series data. An anomaly may be detected according to the anomaly score and a result that indicates a detected anomaly may be provided.

Abstract: Object detection architectures for detecting and classifying objects in an image are modified to incorporate an extending Rapid Class Augmentation (XRCA) progressive learning algorithm with its defining aspect of memory built into its optimizer which allows joint optimization over both the old and the classes using just the new class data and eliminates the issues associated with catastrophic forgetting.

Abstract: Vehicle perception techniques include obtaining a training dataset represented by N training histograms, in an image feature space, corresponding to N training images, K-means clustering the N training histograms to determine K clusters with respective K respective cluster centers, wherein K and N are integers greater than or equal to one and K is less than or equal to N, comparing the N training histograms to their respective K cluster centers to determine maximum in-class distances for each of K clusters, applying a deep neural network (DNN) to input images of the set of inputs to output detected/classified objects with respective confidence scores, obtaining adjusted confidence scores by adjusting the confidence scores output by the DNN based on distance ratios of (i) minimal distances of input histograms representing the input images to the K cluster centers and (ii) the respective maximum in-class.

Abstract: A method for using distributed representations of data items within a first set of data documents clustered in a first two-dimensional metric space to generate a cluster of distributed representations in a second two-dimensional metric space includes clustering in a first two-dimensional metric space, by a reference map generator, a set of data documents, generating a semantic map. A parser generates an enumeration of data items occurring in the set of data documents. A representation generator generates a distributed representation using occurrence information about each data item. A sparsifying module receives an identification of a maximum level of sparsity and reduces a total number of set bits within the distributed representation. The reference map generator clusters, in a second two-dimensional metric space, a set of SDRs retrieved from the SDR database and selected according to a second at least one criterion, generating a second semantic map.

Abstract: A computer-implemented method for controlling a particular computer to execute a task is described. The method includes receiving a control input comprising a visual input, the visual input including one or more screen frames of a computer display that represent at least a current state of the particular computer; processing the control input using a neural network to generate one or more control outputs that are used to control the particular computer to execute the task, in which the one or more control outputs include an action type output that specifies at least one of a pointing device action or a keyboard action to be performed to control the particular computer; determining one or more actions from the one or more control outputs; and executing the one or more actions to control the particular computer.

Abstract: Methods and systems are provided for implementing training of learning models, including obtaining a pre-trained weight set for a learning model on a sample dataset and on a first loss function; selecting at least two tasks having heterogeneous features to be computed by a reference model; obtaining a reference dataset for the at least two tasks; designating a second loss function for feature embedding between the heterogeneous features of the at least two tasks; training the learning model on the first loss function and training the reference model on the second loss function, in turn; and updating the weight set based on a feature embedding learned by the learning model and a feature embedding learned by the reference model, in turn. Methods and systems of the present disclosure may alleviate computational overhead incurred by executing the learning model and loading different weight sets at a central network of the model.

Abstract: A method includes receiving interaction data that indicates, for each given interaction among multiple interactions that occurred at a client device, (i) an event type an (ii) a delay period specifying an amount of time between the given event and a previous event that occurred prior to the given event, encoding each given interaction into an encoded interaction having a standardized format that is a combination of (i) the event type and (ii) the delay period, generating an interaction signature that includes sequence of encoded interactions, processing the sequence of encoded interactions using a model trained to label sequences of user interactions as valid or invalid, including labelling, using the model, a sequence of encoded interactions as invalid, and preventing distribution of a set of content to an entity that performed the sequence of encoded interactions in response to a subsequently identified request to provide content to the entity.

Abstract: Techniques for securing displayed data on computing devices are disclosed. One example technique includes upon determining that the computing device is unlocked, capturing and analyzing an image in a field of view of the camera of the computing device to determine whether the image includes a human face. In response to determining that the image includes a human face, the technique includes determining facial attributes of the human face in the image via facial recognition and whether the human face is that of an authorized user of the computing device. In response to determining that the human face is not one of an authorized user of the computing device, the technique includes converting user data on the computing device from an original language to a new language to output on a display of the computing device, thereby securing the displayed user data even when the computing device is unlocked.

Abstract: Systems and methods for aggregating data. The system is configured to receive metadata from an interactive graphical user interface (GUI) of a user device, aggregate field values from the data stored on one or more databases based on the received metadata and generate filter instructions based on the received metadata. The system is further configured to transmit the aggregated field values and the filter instructions to the user device, receive a user-customized filter set and subscription request for a synthetic symbol associated with the user-customized filter set from the user device, and create the synthetic symbol responsive to the subscription request. Moreover, the system aggregates one or more data values from the data stored on the databases associated with the created synthetic symbol and generates instructions to display the data values on the interactive GUI in accordance with the user-customized filter set associated with the created synthetic symbol.

Abstract: Various embodiments describe techniques for making inferences from graph-structured data using graph convolutional networks (GCNs). The GCNs use various pre-defined motifs to filter and select adjacent nodes for graph convolution at individual nodes, rather than merely using edge-defined immediate-neighbor adjacency for information integration at each node. In certain embodiments, the graph convolutional networks use attention mechanisms to select a motif from multiple motifs and select a step size for each respective node in a graph, in order to capture information from the most relevant neighborhood of the respective node.

Abstract: A mechanism is described for facilitating misuse index for explainable artificial intelligence in computing environments, according to one embodiment. A method of embodiments, as described herein, includes mapping training data with inference uses in a machine learning environment, where the training data is used for training a machine learning model. The method may further include detecting, based on one or more policy/parameter thresholds, one or more discrepancies between the training data and the inference uses, classifying the one or more discrepancies as one or more misuses, and creating a misuse index listing the one or more misuses.

Abstract: A system is provided for providing action inducers to target consumer entities. The system accesses transactions relating to actions involving a consumer entity and a provider entity. The system generates training data based on the transactions. The training data includes a feature vector for each consumer entity that includes values for features derived from the transactions involving that consumer entity and a provider entity. Each feature vector is labeled with an outcome that indicates whether a previously provided action inducer induced the consumer entity to take an action. The system trains a classifier based on the training data, generates a target feature vector for a target consumer entity, applies the classifier to the target feature vector to generate a predicted outcome for the target action inducer, and when the predicted outcome satisfies an inducement criterion, provides the target action inducer to the target consumer entity.

Abstract: A circuit for performing neural network computations for a neural network is described. The circuit includes plurality of neural network layers each including a crossbar arrays. The plurality of crossbar arrays are formed in a common substrate in a stacked configuration. Each crossbar array includes a set of crosspoint devices. A respective electrical property of each of the crosspoint devices is adjustable to represent a weight value that is stored for each respective crosspoint device. A processing unit is configured to adjust the respective electrical properties of each of the crosspoint devices by pre-loading each of the crosspoint devices with a tuning signal. A value of the turning signal for each crosspoint device is a function of the weight value represented by each respective crosspoint device.

Abstract: A method of training a deep neural network, such as would be used in facial recognition, includes training the deep neural network to normalize feature vectors to a learned value representing the radius of a multi-dimensional hypersphere using a convex augmentation of the primary loss function.

Abstract: A method for designing a channel estimation and data detection networks is provided herein. The problem of channel estimation for linear systems has effectively been solved—not the case for non-linear systems. A deep learning framework for channel estimation, data detection, and pilot signal design is described to address the nonlinearity in such systems.

Abstract: A processor-implemented neural network operating method, the operating method comprising obtaining a neural network pre-trained in a source domain and a first style feature of the source domain, extracting a second style feature of a target domain from received input data of the target domain, using the neural network, performing domain adaptation of the input data, by performing style matching of the input data based on the first style feature of the source domain and the second style feature of the target domain, and processing the style-matched input data, using the neural network.

Abstract: A method for controlling a bias of a neural network, the method may include training the neural network by using a loss function that is responsive to classes of a classification process and sensitivity input values thereby setting the bias.

Abstract: One example method includes data protection operations including cyber security operations, threat detection operations, and other security operations. Normal device behavior is learned based on data collected by an anomaly detection engine operating in a kernel. The normal data is used to train a machine learning model. Threats are detected when the machine learning model indicates that new data points deviate from normal device behavior. Associated processes are stopped. This allows threats to be detected based on normal behavior rather than on unknown threat behavior.

Abstract: A reference map generator clusters, into a semantic map, a set of data documents selected according to at least one criterion and associated with a medical diagnosis. A parser generates an enumeration of measurements occurring in the set of data documents. A representation generator generates for each measurement in the enumeration, a sparse distributed representation (SDR). The method includes storing, by a processor on a second computing device, in each of a plurality of memory cells on the second computing device, one of the generated SDRs. A diagnosis support module receives a document comprising a plurality of measurements. The representation generator generates a compound SDR for the document. Each of the plurality of bitwise comparison circuits determine a level of overlap between the compound SDR and the stored generated SDR. The diagnosis support module provides an identification of the medical diagnosis associated with a stored SDR.

Abstract: An audio description system includes a memory and a processor. The memory stores source media comprising frames positioned within the source media according to a time index. The processor is configured to generate, using an image-to-text model, a textual description of each frame; identify intervals within the time index, each interval encompassing one or more positions of one or more frames; identify placement periods within the time index, each placement period being temporally proximal to an interval; generate a summary description based on at least one textual description of at least one frame positioned within a selected interval temporally proximal to a placement period; and associate the summary description with the placement period.

Abstract: The present invention concerns a method and a system for detecting malfunctions in an apparatus and/or defects in a workpiece processed by said apparatus. The method provides for acquiring a sound signal emitted by an apparatus during an operation cycle of the same, then comparing the sound signal with a plurality of audio tracks stored in a memory area for determining a malfunction of the apparatus and/or a defectiveness of the workpiece processed by the apparatus based on the result of said comparison. The operation cycle is subdivided into a plurality of work phases and during the acquisition of the sound signal a work phase of said plurality of work phases is identified. Each audio track of the plurality of stored audio tracks comprises an audio component relating to acquired sound signals, and additional information data comprising at least one identifier of the work phase executed by the apparatus during the acquisition of the sound signals of the audio component.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for prediction of an outcome related to an environment. In one aspect, a system comprises a state representation neural network that is configured to: receive an observation characterizing a state of an environment being interacted with by an agent and process the observation to generate an internal state representation of the environment state; a prediction neural network that is configured to receive a current internal state representation of a current environment state and process the current internal state representation to generate a predicted subsequent state representation of a subsequent state of the environment and a predicted reward for the subsequent state; and a value prediction neural network that is configured to receive a current internal state representation of a current environment state and process the current internal state representation to generate a value prediction.

Abstract: In one example, an apparatus comprises: a memory array having an array of memory elements arranged in rows and columns, each memory element being configured to store a data element; and a memory access circuit configured to: perform a row write operation to store a first group of data elements at a first row of the array of memory elements; perform a column read operation at a first column of the array of memory elements to obtain a second group of data elements; and perform a column write operation to store a third group of data elements at the first column of the array of memory elements to replace the second group of data elements.

Abstract: A set of material candidates expected to yield materials with target properties can be generated. A subject matter expert's decision indicating accepted and rejected material candidates from the set of material candidates can be received. Based on the subject matter expert's input, a machine learning model can be trained to replicate the subject matter expert's decision.

Abstract: A semiconductor device with a small circuit area and low power consumption is provided. The semiconductor device includes first to fourth cells, a current mirror circuit, and first to fourth wirings, and the first to fourth cells each include a first transistor, a second transistor, and a capacitor. In each of the first to fourth cells, a first terminal of the first transistor is electrically connected to a first terminal of the capacitor and a gate of the second transistor. The first wiring is electrically connected to first terminals of the second transistors in the first cell and the second cell, the second wiring is electrically connected to first terminals of the second transistors in the third cell and the fourth cell, the third wiring is electrically connected to second terminals of the capacitors in the first cell and the third cell, and the fourth wiring is electrically connected to second terminals of the capacitors in the second cell and the fourth cell.

Abstract: Multi-task deep learning method for a neural network for automatic pathology detection, comprising the steps: receiving first image data (I) for a first image recognition task; receiving (S2) second image data (V) for a second image recognition task; wherein the first image data (I) is of a first datatype and the second image data (V) is of a second datatype, different from the first datatype; determining (S3) first labeled image data (IL) by labeling the first image data (I) and determining second synthesized labeled image data (ISL) by synthesizing and labeling the second image data (V); training (S4) the neural network based on the received first image data (I), the received second image data (V), the determined first labeled image data (IL) and the determined second labeled synthesized image data (ISL); wherein the first image recognition task and the second image recognition task relate to a same anatomic region where the respective image data is taken from and/or relate to a same pathology to be recogni

Abstract: Techniques for context-based display of content and alert based on distance on a multi-display system are described. For instance, the described techniques can be implemented to determine a device context based on one or more of a first instance of media content displayed on a first display device positioned at a first surface of a client device or an environment in which the client device is positioned. Based on the device context, the described techniques enable output of a second instance of media content via a second display device positioned at a second surface of the client device.

Abstract: Implementations can provide a method that includes: accessing the source code of a script hosted by a remote server; extracting features from the source code in accordance with a machine-learning model comprising one or more layers of logic; at least based on the machine-learning model, determining, for each of the extracted features, a corresponding probability conditioned on the source code containing ransomware; and at least based on the machine-learning model, determining a combined probability for the extracted features conditioned on the source code containing ransomware when the extracted features are jointly present; comparing the combined probability with a threshold; in response to determining that the combined probability exceeds the threshold, flagging the source code as containing ransomware; and in response to determining that the combined probability does not exceed the threshold, flagging the source code as not containing ransomware.

Abstract: A generator for generating artificial data, and training for the same. Data corresponding to a first label is altered within a reference labeled data set. A discriminator is trained based on the reference labeled data set to create a selectively poisoned discriminator. A generator is trained based on the selectively poisoned discriminator to create a selectively poisoned generator. The selectively poisoned generator is tested for the first label and tested for the second label to determine whether the generator is sufficiently poisoned for the first label and sufficiently accurate for the second label. If it is not, the generator is retrained based on the data set including the further altered data. The generator includes a first ANN to input first information and output a set of artificial data that is classifiable using a first label and not classifiable using a second label of the set of labeled data.

Abstract: A method for generating a dual-class dataset is disclosed. A single-class dataset and a context dataset are obtained. The context dataset can be labeled. A model can be trained using the combination of the single-class dataset and the labeled context dataset. The model can be run on the context dataset. The data points that are classified the same as the data points included in the single-class dataset, can be removed from the labeled context dataset and added to the single-class dataset. These steps can be repeated until no data points are classified by the model.

Abstract: A device receives a query from a user associated with a car dealership and applies the query to a first trained machine learning model configured to predict an intent, and to a second trained machine learning model to predict a set of entities. The device generates a normalized representation of the first query that is database language agnostic based on the predicted intent and the predicted set of entities, and translates the normalized representation into a second query having a format compatible with a language of a database of the car dealership. The device fetches data from the database of the car dealership using the second query, and provides the data for display to the user.

Abstract: Systems and methods for predicting yearly performance improvement rates for nearly all definable technologies for the first time are provided. In one embodiment, a correspondence of all patents within the U.S. patent system to a set of technology domains is created. From the identified patent sets, the invention may calculate average centrality of the patents in each domain to predict improvement rates, following a patent network-based methodology. Also disclosed is a system to intake a user technology search query and match user intent with the technology domain as well as the corresponding improvement rate.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating a synthesized signal. A computer-implemented system obtains generator input data including an input signal having one or more first characteristics, processes the generator input data to generate output data including a synthesized signal having one or more second characteristics using a generator neural network, and outputs the synthesized signal to a device. The generator neural network is trained, based on a plurality of training examples, with a discriminator neural network. The discriminator neural network is configured to process discriminator input data that combines a discriminator input signal having the one or more second characteristics with at least a portion of generator input data to generate a prediction of whether the discriminator input signal is a real signal or a synthesized signal.