Abstract: A system for generating an adversarial example in respect of a neural network, the adversarial example generated to improve a margin defined as a distance from a data example to a neural network decision boundary. The system includes a data receiver configured to receive one or more data sets including at least one data set representing a benign training example (x); an adversarial generator engine configured to: generate, using the neural network, a first adversarial example (Adv1) having a perturbation length epsilon1 against x; conduct a search in a direction (Adv1-x) using the neural network; and to generate, using the neural network, a second adversarial example (Adv2) having a perturbation length epsilon2 based at least on an output of a search in the direction (Adv1-x).

Abstract: Methods and apparatuses for accurately determining a model, which is to be the basis of transfer learning, among a plurality of source models, are provided. According to an embodiment, an apparatus for determining a base model to be used for transfer learning to a target domain is provided. The apparatus comprises a memory which comprises one or more instructions and a processor which executes the instructions to construct a neural network model for measuring suitability of a plurality of pre-trained source models, measure the suitability of each of the source models by inputting data of the target domain to the neural network model, and determine the base model to be used for the transfer learning among the source models based on the suitability.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for implementing long-short term memory layers with compressed gating functions. One of the systems includes a first long short-term memory (LSTM) layer, wherein the first LSTM layer is configured to, for each of the plurality of time steps, generate a new layer state and a new layer output by applying a plurality of gates to a current layer input, a current layer state, and a current layer output, each of the plurality of gates being configured to, for each of the plurality of time steps, generate a respective intermediate gate output vector by multiplying a gate input vector and a gate parameter matrix. The gate parameter matrix for at least one of the plurality of gates is a structured matrix or is defined by a compressed parameter matrix and a projection matrix.

Abstract: Disclosed are a method for generating a neural network, an apparatus thereof, and an electronic device. The method includes: obtaining an optimal neural network and a worst neural network from a neural network framework by using an evolutionary algorithm; obtaining an optimized neural network from the optimal neural network by using a reinforcement learning algorithm; updating the neural network framework by adding the optimized neural network into the neural network framework and deleting the worst neural network from the neural network framework; and determining an ultimately generated neural network from the updated neural network framework. In this way, a neural network is optimized and updated from a neural network framework by combining the evolutionary algorithm and the reinforcement learning algorithm, thereby automatically generating a neural network structure rapidly and stably.

Abstract: Provided are an optimization system and method of a deep learning model. According to example embodiments, by optimizing the structure of the deep learning model appropriately for a target dataset without fixing the structure of the deep learning model, it is possible to generate a model structure capable of having high performance on the target dataset and also saving resources.

Abstract: A model generation method includes updating, by at least one processor, a weight matrix of a first neural network model at least based on a first inference result obtained by inputting, to the first neural network model which discriminates between first data and second data generated by using a second neural network model, the first data, a second inference result obtained by inputting the second data to the first neural network model, and a singular value based on the weight matrix of the first neural network model. The model generation method also includes at least based on the second inference result, updating a parameter of the second neural network model.

Abstract: In one aspect, a computer implemented method for efficient value lookup in a set of scalar intervals is provided. The method includes determining, in response to a query for a scalar value, that the scalar value is located in a set of scalar intervals, wherein each of the scalar intervals comprises a left bound and a right bound. The method further includes sorting the scalar intervals based on left bounds. The method further includes comparing, in response to the sorting, a pair of scalar intervals to determine if the pair of scalar intervals overlaps. The method further includes identifying, based on the comparing indicating that the pair overlaps, a method of processing the scalar intervals.

Abstract: A method of constraining data represented in a deep neural network is described. The method includes determining an initial shifting specified to convert a fixed-point input value to a floating-point output value. The method also includes determining an additional shifting specified to constrain a dynamic range during converting of the fixed-point input value to the floating-point output value. The method further includes performing both the initial shifting and the additional shifting together to form a dynamic, range constrained, normalized floating-point output value.

Abstract: An illustrative embodiment includes a method for improving performance of a computer. The method includes: automatically identifying an algorithm supplied by a user for execution on the computer; searching a database of algorithms for at least one algorithm similar to the user-supplied algorithm; determining whether the at least one similar algorithm will improve performance of the computer relative to the user-supplied algorithm; and if the at least one similar algorithm will improve performance of the computer relative to the user-supplied algorithm, modifying the user-supplied algorithm to incorporate at least in part the at least one similar algorithm.

Abstract: An artificial intelligence (AI) system that utilizes a machine learning algorithm, such as deep learning, etc. and an application of the AI system is provided. A method, performed by a server, of integrating and managing a plurality of databases (DBs) includes obtaining a plurality of knowledge graphs related to DBs generated from the plurality of DBs having different structures from one another, inputting the plurality of knowledge graphs related to DBs into a learning model related to DB for determining a correlation between data in the plurality of DBs, and obtaining a virtual integrated knowledge graph output from the learning model related to DB and including information about a correlation extracted from the plurality of knowledge graphs related to DBs.

Abstract: Historical behavioral information of a user is retrieved, where the historical behavioral data includes data associated to operations performed by the user on a server. Recommended information sets are determined based on the historical behavioral information. A plurality of weight coefficients are generated for the plurality of recommended information sets. A recommendation list is determined based on the plurality of weight coefficients. It is determined whether the recommendation list satisfies a recommendation condition. If the recommendation list satisfies the recommendation condition, a recommendation based on the recommendation list is transmitted to the user device.

Abstract: Systems, methods and computer program products for forecast data storage. Embodiments implement fine-grained forecast data management. A cloud-based object storage system capable of storing multiple versions of an object in a container is identified. A forecast data set covering a relatively longer time period (e.g., years) is partitioned into fine-grained forecast data items corresponding to relatively shorter forecast data time periods (e.g., months, days). Some of the fine-grained forecast data items corresponding to the relatively shorter forecast data time periods are stored into a first portion of metadata of the container rather than storing the forecast data items into the object itself. Updated variations of the fine-grained forecast data items and/or new forecast data items are stored in versions of the object. A second portion of metadata of the container is used to describe a version mapping between the forecast data time periods and corresponding object versions in the container.

Abstract: An apparatus and method for detecting impact factors for an operating environment. The apparatus generates a detection result for each of the first factors of a plurality of first historical records by analyzing a dissimilarity degree of the plurality of first data corresponding to each first factor. Each detection result is a continuous data type or a discrete data type. The apparatus trains a data type recognition model according to the first historical records and the detection results. The apparatus establishes a basic prediction model by a training set of a plurality of second historical records, generates a comparison set by rearranging the second data corresponding to a specific factor in the training set, establishes a comparison prediction model by the comparison set, and determines a degree of importance of the specific factor by comparing the accuracies of the basic prediction model and the comparison prediction model.

Abstract: A method includes detecting a defective entigen group within a knowledge database. The defective entigen group includes entigens and one or more entigen relationships between at least some of the entigens. The defective entigen group represents knowledge of a topic. The method further includes obtaining corrective content for the topic based on the defective entigen group and generating a corrective entigen group based on the corrective content. The method further includes updating the defective entigen group utilizing the corrective entigen group to produce a curated entigen group.

Abstract: The system can include a rules engine and one or more application systems. The rules engine can be configured to perform receiving overrides, storing the overrides in an overrides repository, generating a bloom filter using the overrides, and sending the bloom filter to the one or more application systems. The one or more application systems can be configured to perform storing the bloom filter as a cached bloom filter, receiving a request to evaluate rules and check for the overrides, and determining, using the cached bloom filter, whether to apply any of the overrides to the request.

Abstract: The present disclosure relates to a method for enabling data integration. The method comprises collecting matching results of matching of records by a matching component over a time window. The number of false tasks of user defined tasks and system defined tasks in the collected matching results may be determined. The matching criterion used by the matching component may be adjusted to minimize the number of user defined tasks while the fraction of false tasks stays within a certain limit. The matching criterion may be replaced by the adjusted matching criterion for further usage of the matching component.

Abstract: The techniques herein include using an input context to determine a suggested action and/or cluster. Explanations may also be determined and returned along with the suggested action. The explanations may include (i) one or more most similar cases to the suggested case (e.g., the case associated with the suggested action) and, optionally, a conviction score for each nearby cases; (ii) action probabilities, (iii) excluding cases and distances, (iv) archetype and/or counterfactual cases for the suggested action; (v) feature residuals; (vi) regional model complexity; (vii) fractional dimensionality; (viii) prediction conviction; (ix) feature prediction contribution; and/or other measures such as the ones discussed herein, including certainty. The explanation data may be used to determine whether to perform a suggested action.

Abstract: Embodiments generate digital plans for agricultural fields. In an embodiment, a model receives digital inputs including stress risk data, product maturity data, field location data, planting date data, and/or harvest date data. The model mathematically correlates sets of digital inputs with threshold data associated with the stress risk data. The model is used to generate stress risk prediction data for a set of product maturity and field location combinations. In a digital plan, product maturity data or planting date data or harvest date data or field location data can be adjusted based on the stress risk prediction data. A digital plan can be transmitted to a field manager computing device. An agricultural apparatus can be moved in response to a digital plan.

Abstract: An example computer-implemented method includes receiving agricultural data records comprising a first set of yield properties for a first set of seeds grown in a first set of environments, and receiving genetic feature data related to a second set of seeds. The method further includes generating a second set of yield properties for the second set of seeds associated with a second set of environments by applying a model using the genetic feature data and the agricultural data records. In addition, the method includes determining predicted yield performance for a third set of seeds associated with one or more target environments by applying the second set of yield properties, and generating seed recommendations for the one or more target environments based on the predicted yield performance for the third set of seeds. In the present example, the method also includes causing display, on a display device communicatively coupled to the server computer system, the seed recommendations.

Abstract: According to one embodiment, An apparatus for preprocessing a security log includes a field divider configured to divide a character string of a security log into a plurality of fields on the basis of a structure of the security log, an ASCII code converter configured to convert a character string included in each of the plurality of divided fields into ASCII codes, and a vector data generator configured to generate vector data for each of the plurality of divided fields using the converted ASCII codes.

Abstract: Methods for recommending next user input using pattern analysis of user input is provided. According to an aspect of the present disclosure, a method comprising obtaining information on a series of user inputs entered through a graphic user interface (GUI), analyzing the information on the series of user inputs to identify a pattern formed by the series of user inputs, and when the pattern is identified, automatically displaying next input recommendation information determined depending on the identified pattern and a last user input of the series of user inputs without additional user input after the series of user inputs, is provided.

Abstract: A predictive modeling method may include obtaining a fitted, first-order predictive model configured to predict values of output variables based on values of first input variables; and performing a second-order modeling procedure on the fitted, first-order model, which may include: generating input data including observations including observed values of second input variables and predicted values of the output variables; generating training data and testing data from the input data; generating a fitted second-order model of the fitted first-order model by fitting a second-order model to the training data; and testing the fitted, second-order model of the first-order model on the testing data. Each observation of the input data may be generated by (1) obtaining observed values of the second input variables, and (2) applying the first-order predictive model to corresponding observed values of the first input variables to generate the predicted values of the output variables.

Abstract: A client device determines a local user gradient value based on a current user preference vector and a local item gradient value based on a current item feature vector. The client device updates a user preference vector by using the local user gradient value and updates an item feature vector by using the local item gradient value. The client device determines a neighboring client device based on a predetermined adjacency relationship. The local item gradient value is sent by the client device to the neighboring client device. The client device receives a neighboring item gradient value sent by the neighboring client device. The client device updates the item feature vector by using the neighboring item gradient value. In response to the client device determining that a predetermined iteration stop condition is satisfied, the client device outputs the user preference vector and the item feature vector.

Abstract: Systems and methods are provided in relation to a complex adaptive command guided swarm system that includes an operator section comprising a first command and control section and a plurality of networked swarm of semi-autonomously agent controlled system of systems platforms (SAASoSP). The first command and control section includes a user interface, computer system, network interface, a plurality of command and control systems executed or running on the computer system.

Abstract: Customer relationship management (“CRM”) implemented in a computer system, including parsing, by a parsing engine of the computer system into parsed triples of a description logic, words of a CRM event from an incoming stream of CRM events, the CRM event characterized by an event type, the stream implemented in a CRM application of the computer system; and inferring, by an inference engine from the parsed triples according to inference rules specific to the event type, inferred triples.

Abstract: Systems and methods for detecting indirect bias in machine learning models are provided. A computer-implemented method includes: receiving, by a computer device, a user request to detect transitive bias in a machine learning model; determining, by the computer device, correlations of attributes of neighboring data not included in a dataset of the machine learning model; ranking, by the computer device, the attributes based on the determined correlations; and returning, by the computer device, a list of the ranked attributes to a user that generated the user request.

Abstract: Systems and methods for vehicle simulation are provided. A method can include obtaining generator input data indicative of one or more parameter values, and inputting the generator input data into a machine-learned generator model that is configured to generate artificial data based at least in part on the generator input data. The artificial data can include data representing an artificial scenario associated with an autonomous vehicle. The method can include obtaining an output of the machine-learned generator model that can include the artificial data, and inputting the artificial data into a machine-learned discriminator model to generate authenticity data representing an authenticity associated with the artificial scenario of the artificial data. The method can include obtaining an output of the machine-learned discriminator model that can include the authenticity data. The method can include selecting the artificial scenario in the artificial data.

Abstract: Online learning of model parameters is performed by obtaining a first target value in a target sequence and a feature vector corresponding to the first target value. The feature vector includes a plurality of elements. The feature vector can be modified to obtain a modified feature vector by reducing an absolute value of at least one element of the feature vector. An inverse Hessian matrix can be generated recursively from a previous inverse Hessian matrix using at least the feature vector and the modified feature vector. Parameters of a model can be updated using the inverse Hessian matrix.

Abstract: Example implementations relate to classifying parts. A computing device may comprise a processing resource; and a memory resource storing non-transitory machine-readable instructions to cause the processing resource to: receive a part description of a part; classify the part by determining a commodity of the part based on the part description using machine learning; and update attributes of the part based on the determined commodity of the classified part.

Abstract: Methods, systems, and devices for determining device associations are described. Some database systems may store information related to device characteristics. Each of these devices may be operated by one or more users, and each user may operate one or more devices. In some cases, information about users may be more valuable than information about devices. As such, a system may determine probable associations between devices, where an association can correspond to operation by a same user. To determine device associations, the system may perform a machine-learning process (e.g., using probabilistic soft logic (PSL) and a hinge-loss Markov Random Field (HL-MRF) model) on input device characteristics and connection information to generate a probability density function. The probability density function may indicate associations between devices within the system.

Abstract: A computer-implemented method includes employing a dynamic Boltzmann machine (DyBM) to solve a maximum likelihood of generalized normal distribution (GND) of time-series datasets. The method further includes acquiring the time-series datasets transmitted from a source node to a destination node of a neural network including a plurality of nodes, learning, by the processor, a time-series generative model based on the GND with eligibility traces, and, performing, by the processor, online updating of internal parameters of the GND based on a gradient update to predict updated times-series datasets generated from non-Gaussian distributions.

Abstract: A method and system are herein disclosed. The method includes developing a joint latent variable model having a first variable, a second variable, and a joint latent variable representing common information between the first and second variables, generating a variational posterior of the joint latent variable model, training the variational posterior, and performing inference of the first variable from the second variable based on the variational posterior.

Abstract: In one aspect, a computer implemented method for translating and executing rules using a directed acyclic graph is provided. The method includes transforming a ruleset into a directed acyclic graph. The directed acyclic graph includes a plurality of nodes and a plurality of branches. The method further includes identifying similarities across the plurality of branches. The method further includes grouping branches of the directed acyclic graph based on the identified similarities. The method further includes creating a modified directed acyclic graph based on the grouping. The method further includes selecting and using a method of processing a group of the modified directed acyclic graph based on an aspect of the group.

Abstract: Techniques are provided for improving quantum circuits. The technology includes approximately expanding, by a system operatively coupled to a processor, using zero to a number of applications of a super controlled basis gate, a target two-qubit operation, with the approximately expanding resulting in instances of the target two-qubit operation corresponding to the zero to the number of applications, and the target two-qubit operation is part of a source quantum circuit associated with a quantum computer. The system analyzes the instances and the super controlled basis gate, and automatically rewrites the source quantum circuit into a deployed quantum circuit based on the analyzing.

Abstract: A hybrid quantum-classical (HQC) computer prepares a quantum Boltzmann machine (QBM) in a pure state. The state is evolved in time according to a chaotic, tunable quantum Hamiltonian. The pure state locally approximates a (potentially highly correlated) quantum thermal state at a known temperature. With the chaotic quantum Hamiltonian, a quantum quench can be performed to locally sample observables in quantum thermal states. With the samples, an inverse temperature of the QBM can be approximated, as needed for determining the correct sign and magnitude of the gradient of a loss function of the QBM.

Abstract: In various embodiments, A quantum computing system comprises at least one classical processor configured by operational instructions stored in a classical memory to perform operations including: generating a qubit encoding from an input, wherein the qubit encoding indicates a sublattice of a projective Riemann-hypersphere that represents a plurality of qubits; and generating a quantum solution based on quantum calculations wherein the quantum calculations include a decomposition of the sublattice performed via a plurality of iterations.

Abstract: A system of training behavior labeling model is provided. Specifically, a processing unit inputs each data of a training data set into a plurality of learning modules to establish a plurality of labeling models. The processing unit obtains a plurality of second labeling information corresponding to each data of a verification data set and generates a behavior labeling result according to the second labeling information corresponding to each data of the verification data set. The processing unit obtains a labeling change value according to the behavior labeling result and first labeling information corresponding to each data of the verification data set. The processing unit, if determining that the labeling change value is greater than a change threshold, updates the first labeling information according to the behavior labeling results, exchanges the training data set and the verification data set and reestablishes the labeling models.

Abstract: A tendency of an action of a robot may vary based on learning data used for training. The learning data may be generated by an agent performing an identical or similar task to a task of the robot. An apparatus and method for updating a policy for controlling an action of a robot may update the policy of the robot using a plurality of learning data sets generated by a plurality of heterogeneous agents, such that the robot may appropriately act even in an unpredicted environment.

Abstract: A method of training a student model corresponding to a teacher model is provided. The teacher model is obtained through training by taking first input data as input data and taking a corresponding output data as an output target. The method comprises training the student model by taking second input data as input data and taking the corresponding output data as an output target. The second input data is data obtained due to changing of the first input data.

Abstract: A distribution system 100 includes a data management apparatus 10 and a plurality of calculators 20 that execute machine learning. The data management apparatus 10 includes a data acquisition unit 11 that acquires information regarding training data held in a memory 21 of each of the calculators 20, from the calculators 20, and a data rearrangement unit 12 that determines training data that is to be held in the memory 21 of each of the calculators 20, based on characteristics of the machine learning processes that are executed by the calculators 20, and the information acquired from the calculators.

Abstract: A method, device and computer program product for deep learning are provided. According to one example, a parameter related to a deep learning model for a training dataset allocated to a server is obtained at a client; a transmission state of the parameter is determined, the transmission state indicating whether the parameter has been transmitted to the server; and information associated with the parameter to be sent to the server is determined based on the transmission state to update the deep learning model. Therefore, the performance of deep learning may be improved, and the network load of deep leaning may be reduced.

Abstract: A retail agent determines a parameter of an activity proposal model by using data stored in a past record database, and determines parameters of an activity determination model and an activity value evaluation model by further using base activity simulation data. Consequently, it is possible to appropriately determine parameters of a subsystem control method in a complex system which cannot be embodied as a simulator and shows a significant change with respect to past record data.

Abstract: A method includes performing a first clustering operation to group members of a first data set into a first group of clusters and associating each cluster of the first group of clusters with a corresponding label of a first group of labels. The method includes performing a second clustering operation to group members of a combined data set into a second group of clusters. The combined data set includes a second data set and at least a portion of the first data set. The method includes associating one or more clusters of the second group of clusters with a corresponding label of the first group of labels and generating training data based on a second group of labels and the combined data set. The method includes training a machine learning classifier based on the training data to provide labels to a third data set.

Abstract: One or more embodiments are directed to identifying documents with topic vectors by training a machine learning model with a training documents generated from text collections, receiving, after generating a list of topic vectors for the plurality of text collections, an additional text collection, and generating an additional topic vector for the additional text collection without training the machine learning model on the additional text collection. One or more embodiments further include updating the list of topic vectors with additional topic vectors that includes the additional topic vector, receiving a first topic vector based on a first text collection generated in response to user interaction, and matching the first topic vector to the additional topic vector. One or more embodiments further include presenting a link corresponding to the additional text collection in response to matching the first topic vector to the additional topic vector.

Abstract: Rapid knowledge base discovery techniques. A database describes associations between knowledge base articles and closed problem cases. In periodic batch operations, the associations are used to generate solution probability predictors, each of which predictor corresponds to a probability that a particular knowledge base article was used to resolve a particular problem or case. The solution probability predictors comprise probability predictor parameter values associated with the set of words that occur in closed customer problem cases. A specialized data structure is populated with the probability predictor parameter values. When a new active customer problem case is opened, a set of words pertaining to the new, active case is constructed. The active case words are used with the specialized data structure to generate a probability value for each of a set of knowledge base articles. The knowledge base articles having the highest probability values are identified and presented in an ordered list.

Abstract: Systems and methods are provided for receiving a request for services in a given location from a client device operated by a user and generating a set of features based on information included in the request for services in the given location. The systems and methods further provide for analyzing the set of features using a machine learning model to predict whether only services that can be instantly booked should be provided in response to the request for services in the given location, analyzing a prediction output by the machine learning model to determine that only services that can be instantly booked should be provided in response to the request for services in the given location, and generating a list with only services that can be instantly booked.

Abstract: Provided are a system and method for determining whether an apparent booking is a genuine booking or is a blocked period of unavailability that is not the result of a genuine booking. Bookings occur in all sorts of industries, such as travel, medical, entertainment, weddings, catering, and the like. In some examples, the method may include receiving content from a website that includes a listing for an object, identifying a period of unavailability of the object based on the content received from the website, predicting, via a machine learning model, whether the period of unavailability of the object is a blocked period that is not a result of a reservation of the object, the predicting being performed based on additional content visible on the website being input into the machine learning model, and storing an identifier of the period of unavailability and information about the prediction within a storage device.

Abstract: A cargo management system for handling cargo transportation is disclosed. The management system can be configured to obtain information regarding one or more alternative routes of a journey of a cargo-shipping-unit (CSU). The system comprises an operational recommending unit (ORU), which is configured to obtain one or more define targets that are related to the type of cargo that is carried by the CSU and to present the impact of each alternative route on each of the define targets.

Abstract: A method for asset management of power equipment compensates a reference reliability model by each sub-device of the power equipment and generate a unique reliability model by the each sub-device by comparing reliability of the reference reliability model by the each sub-device and the health index by the each sub-device; assesses priorities based on equipment sensitivity and establishes a maintenance strategy while analyzing reliability by substation system reliability and reliability by economic value; calculates reliability of the power equipment by applying a system relationship model between the power equipment and the each sub-device to which specific weight and failure rate are reflected; selects a maintenance scenario by a predetermined priority; and updates the reliability model for the power equipment and the unique reliability model by the each sub-device as a result of executing maintenance.

Abstract: A method, an apparatus, and a computer program product for accessing electronic medical records are provided in which a portable computing device uniquely associated with a user authenticates an identification of the user and automatically retrieves information corresponding to the user from electronic healthcare records systems using the identification. The retrieved information may be combined with other information and electronically delivered to a healthcare provider.