International Business Machines Corporation

Mitsubishi Materials Corporation

Shoichiro Watanabe

Yasutaka Nishimura

Kenichi Takasaki

Sanehiro Furuichi

Tomohiro Shioya

Satoshi Hosokawa

Hirofumi Nishikawa

Mari Abe Fukuda

Kei Sugano

Takahito Tashiro

Hidehiko Nagaya

Yasuharu Imai

Norio Aso

Makoto Tanibayashi

Hiroya Ogihara

Tateo Kawamura

Jian Ding

Zhi Hu Wang

Masami Tada

Asuka Unno

Asuka Tokunaga

Kenji Ishizawa

Xin Zhang

Ning Duan

Shiwan Zhao

Chun Ma

Yasuhisa Gotoh

Daisuke Maruyama

Keisuke Nitta

Yuka Sasaki

Bang An

Kaoru Hosokawa

Kouichi Kageyama

Nan Xia

Jing Lan Liu

Takuma Kiso

Hiroyuki Norigoe

A computer-implemented method, a computer program product, and a computer system for parallel cross validation in collaborative machine learning. A server groups local models into groups. In each group, each local device uses its local data to validate accuracies of the local models and sends a validation result to a group leader or the server. The group leader or the server selects groups whose variances of the accuracies are not below a predetermined variance threshold. In each selected group, the group leader or the server compares an accuracy of each local model with an average value of the accuracies and randomly selects one or more local models whose accuracies do not exceed a predetermined accuracy threshold. The server obtains weight parameters of selected local models and updates the global model based on the weight parameters.

Parallel cross validation in collaborative machine learning

A method for calculating traffic flow changes includes detecting a traffic event. The method further includes determining an affected area of the traffic event and determining an investigation area based on the affected area. The method further includes selecting at least one vehicle located within the investigation area and calculating a change in traffic flow due to the traffic event based on a comparison of a predicted traffic flow with a current traffic flow, wherein the current traffic flow is based on information received from the at least one vehicle. The method further includes updating the affected area based on the change in traffic flow and calculating an updated change in traffic flow based on the updated affected area when the updated affected area is larger than a predetermined threshold area.

Calculating traffic flow changes due to traffic events

Aspects of the invention include receiving an identity of a primary web-based material from a user, wherein the primary web-based material describes a topic. Subparts of the primary web-based material are indexed. Supplemental web-based materials accessed by a plurality of reference users are retrieved, wherein the supplemental web-based materials relate to the subparts. A sequence of supplemental web-based materials is generated for the user to follow to gain an understanding of the topic.

Efficiently manage and share resources during e-learning

A computer system trains a federated learning model. A federated learning model is distributed to a plurality of computing nodes, each having a set of local training data comprising labeled data samples. Statistical data is received from each computing node that indicates the node's count of data samples for each label, and is analyzed to identify one or more computing nodes having local training data in which a label category is underrepresented beyond a threshold value with respect to data samples. Additional data samples labeled with the underrepresented labels are provided, and the computing nodes perform training. Results of training are received and are processed to generate a trained global model. Embodiments of the present invention further include a method and program product for training a federated learning model in substantially the same manner described above.

Training federated learning models

This disclosure provides embodiments for context based vehicular traffic prediction. A trained neural network modeling a relationship between historical traffic data and associated historical contextual data for a roadway link is obtained. Expected contextual data for a future time period for the roadway link is acquired. Predicted traffic data for the future time period for the roadway link is generated with the trained neural network based on the expected contextual data.

Context based vehicular traffic prediction

Utilizing a trained generative adversarial network (GAN) model to cause a computer to output multivariate forecasted time-series data by providing a trained GAN model, the GAN model comprising dilated convolutional layers for receiving time-series multivariate data, receiving time-series multivariable biometric data, generating, using the GAN model, successive time series multivariate biometric data according to the time-series multivariate biometric data, determining an outcome according to the successive time-series multivariate biometric data, and providing an output associated with the outcome.

Forecasting multivariate time series data

A method and/or system for managing a database that stores space-time context objects is provided. The system receives a query range in a multi-dimensional space. The system maps the query range into a set of fragments of a space-filling curve that fills the multi-dimensional space in all dimensions of the multi-dimensional space. The system uses each mapped fragment in the set of mapped fragments as a key to query the database for space-time context objects that are mapped to the space-filling curve. The system queries the database by identifying one or more context objects that intersect the mapped fragment at the space-filling curve.

Spatial-temporal query for cognitive IoT contexts

An embodiment of the invention may include a method, computer program product and computer system for managing mobile objects. The embodiment may receive an event detected in a geographic space by a mobile object of a plurality of mobile objects. The embodiment may determine a reliability of the event based on a reliability associated with the mobile object. The embodiment may store the event in a mobile object database based on the reliability of the even being above a reference reliability.

Management of mobile objects

Embodiments are directed to a user interface for superimposingly displaying, on a map, a plurality of travel route patterns reflecting tendencies of movements of a moving object and a new travel route predicted using the plurality of travel route patterns. The plurality of travel route patterns being extracted in advance by analyzing traveling history data of the moving object and being displayed in different manners, according to a prediction probability between each of the plurality of travel route patterns and the new travel route.

User interface

Described are techniques for acquiring geospatial data according to an information value. The techniques including determining a context for geospatial data to be used in an application, where the context is based on one or more external factors that influence variation of the geospatial data. The techniques further include calculating an information value of the geospatial data in the context for each of a plurality of information acquisition methods, where the plurality of information acquisition methods include respective data acquisition frequencies and respective spatial resolutions, and where the information value is based on an information loss function, an information amount, and a cost. The techniques further include selecting a first information acquisition method with a highest information value and acquiring the geospatial data using the first information acquisition method.

Geospatial data acquisition based on information value

The present invention may be a method, a computer system, and a computer program product for grouping a plurality moving objects capable of communicating with a server computer. The server computer performs the method comprising: predicting travel routes of the plurality of moving objects, using current traveling data and travel history data; grouping the plurality of moving objects into at least one group, using the predicted travel routes; determining a representative moving object in each group; and communicating with the representative moving object.

Grouping of moving objects

Embodiments of the present invention provide a computer system, a computer program product, and a method that comprises in response to receiving a data packet from a computing device, classifying the data packet as a task having one or more portions; allocating the classified task to a processing location within a data region based on a location of the computing device; in response to a change associated with the task, dynamically calculating alternate processing locations within a radius of the data region to process one or more portions of the task based on scoring values associated with the change; and redistributing at least one portion of the classified task according to an alternate processing location of dynamically calculated alternate processing locations.

Dispatching tasks and data using multi-access edge computing

A method, computer system, and a computer program product for supplementing learning data to determine a Most Probable Path (MPP) for a user driver is provided. The present invention may include determining a change associated with a set of map data. The present invention may then include extracting a set of trajectory data in response to the determined change associated with the set of map data. The present invention may also include performing a rerouting calculation for the extracted set of trajectory data, wherein the performed rerouting calculation generates a rerouting result, wherein the generated rerouting result is added to a set of learning data in a learning data database. The present invention may further include adding one or more trajectory patterns to the learning data based on an increase frequency after the determined change associated with the set of map data.

Supplementing learning data to determine most probable path

A method of calculating a most probable path (MPP) comprising: with a classification module, classifying each of a plurality of origin and destination (O/D) trajectories into one of a plurality of periods of time, the classification comprising: changing a data sampling time period of any of the plurality of trajectories within the determined classification periods; with the classification module, detecting any O/D data descriptive of an aperiodic O/D trajectory using context data descriptive of the aperiodic O/D trajectory; and with the classification module, calculating a route trajectory serving as a candidate of the MPP to determine a priority O/D trajectory that matches the context data.

Calculating a most probable path

Disclosed embodiments provide techniques for generation of vehicle travel data. Past-travel data including previously taken trips, along with metadata including, but not limited to, vehicle type, vehicle model, navigation device model, are obtained and classified. A travel route for simulation data is selected. Data from various previous trips are combined to create a simulation data set for the travel route. This data is based on actual vehicle data. This can enable a new level of accuracy for data analysis and simulations based on this data as compared with purely simulated data. Actual vehicle data accounts for driver behavior, road obstacles, signage, vegetation, vehicle characteristics, and other factors that may not be able to be captured with purely simulated traffic data. Using the data generated by disclosed embodiments enables these factors to be considered in traffic analysis.

Growing community of inventors

Tokyo, Japan

Shoichiro Watanabe