International Business Machines Corporation

Shiwan Zhao

Zhong Su

Li Ni Zhang

Honglei Guo

Keke Cai

Zhi Hu Wang

Changhua Sun

Xiaolu Zhang

Dongxu Duan

Quan Yuan

Zhili Guo

Sheng Hua Bao

Shi Lei Zhang

Xian Wu

Bai Chen Deng

Enliang Xu

Jian Ding

Hao Kai Zhang

Yi Ke Wu

Yue Chen

Rong Fu

Xia Tian Zhang

Yi Qin Yu

Wei Gao

Jing Mei

Jing Ding

Song Xu

Guoyu Tang

Elaine M Branagh

Pitipong Jun Sen Lin

Bang An

Bo Yang

Jie Zhang

Jian Ye Chen

Qing Fa Wang

Jim Alain Laredo

Yong Qin

Lin Luo

Yan Yan

Zhen Xiang Yang

Xu Min

Dong Xu Duan

Qining Li

Xi Qing Wang

Meng Ting Hu

Hao Zhang

Marco Pistoia

Tao Wang

Xin Zhang

Birgit Monika Pfitzmann

Peng Gao

Shao Chun Li

Chun Ma

Wen Liu

Min Li

Shoichiro Watanabe

Jian Min Jiang

Guo Tong Xie

Guo Qiang Hu

Li Zhang

Hui Su

Qin Shi

Xiao Ping Guo

Yubo Li

Junsong Wang

Yuan Ni

ZhePeng Wang

Ke Ke Cai

Alessandro Morari

Yunhui Zheng

Ke Jin

Eryu Xia

Wei Hong Qian

Yuan Zhou

Xiu Li Li

Roslyn I Hickson

Guo Yu Tang

Zhi Qiao

Bing Wu

Xin Chang

Yan Fei Wang

Dingli Gao

Bingzhe Wu

Yong Ding

Jing Lan Liu

Chia Yeow Khiang

Embodiments of the present disclosure relate to a method, system, and computer program product for domain adaptation. According to the method, a source model of a source domain is obtained, where the source model is trained to generate a label indicating a predicted category of data from the source domain. A training sample from a target domain is obtained, where the training sample comprises training data from the target domain and a true label indicating a true category of the training data from the target domain. A first label is generated for the training data by using the source model. The first label indicates a predicted category of the training data. A target model of the target domain is trained based on the training data, the true label and the first label.

Domain adaptation

Methods and systems for reinforcement learning with dynamic agent grouping include gathering information at a first agent using one or more sensors. Shared information is received at the first agent from a second agent. An agent model is trained at the first agent using the gathered information and the shared information. A contribution of the shared information is weighted according to a degree of similarity between the first agent and the second agent. An action is generated using the trained agent model responsive to the gathered information.

Reinforcement learning by sharing individual data within dynamic groups

Aspects of the invention include a computer-implemented method including extracting a domain-specific object feature from a first image data, wherein the feature describes an object in the first image data. A domain-specific semantic meaning of text data is determined. The object feature is mapped to a portion of the text data, wherein the portion of the text data describes the object. A joint representation of the object and the portion of the text data is created. A second image data and a query directed towards an object in the second image data is received. An answer to the query is generated based on the joint representation.

Medical visual question answering

In an approach to augmenting a caption dataset by leveraging a denoising autoencoder to sample and generate additional captions from the ground truth captions, one or more computer processors generate a plurality of new captions utilizing an autoencoder fed with one or more noisy captions, wherein the autoencoder is trained with a dataset comprising a plurality of ground truth captions. The one or more computer processors calculate an importance weight for each new caption in the plurality of generated new captions as compared to a plurality of associated ground truth captions based on a consensus metric. The one or more computer processors train a caption model with the generated plurality of new captions and associated calculated weights.

Denoising autoencoder image captioning

Approaches presented herein enable fault detection. More specifically, implementation code of one or more functions is identified from source code. The implementation code of the one or more functions is converted to corresponding Abstract Syntax Trees (ASTs). The implementation code of the one or more functions is represented as a first plurality of sets of AST paths over the ASTs. Classification results for the one or more functions are generated with a classifier based on the first plurality of sets of AST paths for the implementation code of the one or more functions. Each of the classification results indicates a probability of having at least one fault in a corresponding function of the one or more functions. Fault detection results of the source code are generated based on the classification results.

Source code fault detection

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

In an approach to improving the image captioning performance of low-resource languages by leveraging multimodal inputs, one or more computer processors encode an image utilizing an image encoder, wherein the image is contained within a triplet comprising the image, one or more high-resource captions, and one or more low-resource captions. The one or more computer processors generate one or more high-resource captions utilizing the encoded image and the triplet inputted into a high-resource decoder. The one or more computer processors encode the one or more generated high-resource captions utilizing a high-resource encoder. The one or more computer processors add adaptive cycle consistency constraints on a set of calculated attention weights associated the triplet. The one or more computer processors generate one or more low-resource captions by simultaneously inputting the encoded image, the encoded high-resource caption, and the triplet into a trained low-resource decoder.

Adaptive cycle consistency multimodal image captioning

Techniques are provided for training machine learning model. According to one aspect, a training data is received by one or more processing units. The machine learning model is trained based on the training data, wherein the training comprises: optimizing the machine learning model based on stochastic gradient descent (SGD) by adding a dynamic noise to a gradient of a model parameter of the machine learning model calculated by the SGD.

Training machine learning model

A computer implemented method, computer system and computer program product are provided for aspect-based sentiment analysis. According to the method, at least one sentence and at least one aspect are received by one or more processing units. At least one opinion snippet of the at least one sentence is determined based on the at least one sentence and the at least one aspect, through a trained aspect-sentence fusion model, by one or more processing units. A sentiment prediction of the at least one opinion snippet is calculated by one or more processing units for the at least one aspect.

Opinion snippet detection for aspect-based sentiment analysis

A computer-implemented method is provided for allocating a set of shared dockless bikes. The method includes providing a notification to a user of an availability prediction of a shared dockless bike, from among the set of shared dockless bikes, based on a trajectory of the shared dockless bike with respect to a current location of the user, and adding the user to a request queue associated with the current location of the user, responsive to the user agreeing to wait a specified time period for access to the shared dockless bike. The method further includes maintaining an available bikes queue for newly available shared dockless bikes, from among the shared dockless bikes, responsive to the newly available ones of the shared dockless bikes becoming available. The method also includes notifying users of the newly available shared dockless bikes in a priority order determined from the request queue.

Dynamic booking system for shared dockless bikes using trajectory position

A computer-implemented method for predicting non-communicable diseases with infectious risk factors using artificial intelligence includes detecting one or more risk factors associated with a non-communicable disease based on a graph associated with person-to-person links, generating a data structure for compactly representing the graph to compute at least one person-to-person distance, and performing a machine learning technique with regularization of the at least one person-to-person distance.

Predicting non-communicable disease with infectious risk factors using artificial intelligence

Using a natural language processing model, a historical defect report comprising a defect description in narrative text form is parsed. Within a code repository, source code associated with the historical defect report is identified. From the historical defect report and the source code, a logging rule comprising a defect type, logging placement information corresponding to the defect type, and logging format information corresponding to the defect type is generated. By parsing a new defect report using the natural language processing model, the new defect report reporting a defect in new source code, it is determined that the logging rule applies to the new defect report. Logging source code generating logging output when executed is placed within the new source code according to the logging rule. Execution of the new source code including the logging source code is caused, generating the logging output.

Automated defect type based logging integration within source code

In an approach to augmenting caption datasets, one or more computer processors sample a ratio lambda from a probability distribution based on a pair of datapoints contained in a dataset, wherein each datapoint in the pair of datapoints comprises an image and an associated caption; extend the dataset by generating one or more new datapoints based on the sampled ratio lambda for each pair of datapoints in the dataset, wherein the sampled ratio lambda incorporates an interpolation of features associated with the pair of datapoints into the generated one or more new datapoints; identify one or more objects contained within a subsequent image utilizing an image model trained utilizing the extended dataset; generate a subsequent caption for one or more identified objects contained within the subsequent image utilizing a language generating model trained utilizing the extended dataset.

Mixup image captioning

A computer implemented method, computer system and computer program product are provided for data classification. According to the method, an original data is received by one or more processing units. A classification of the original data with a trained classifier is determined by one or more processing units, wherein the classifier is trained by a labeled data of different granularity, and the labeled data include at least one defect object.

Classifier trained with data of different granularity

Aspects of the present invention disclose a method, computer program product, and system for image processing. The method includes one or more processors generating a first set of binary images from a first image based on a first color attribute value range associated with a textual object presented in the first image. The method further includes one or more processors recognizing a first set of candidates for the textual object from the first set of binary images. The method further includes one or more processors determining a first appearance frequency of a first candidate in the first set of candidates. In response to determining that the first appearance frequency exceeds a first frequency, threshold he method further includes one or more processors determining that the first candidate is a first recognition result for the textual object in the first image.

Image recognition

Growing community of inventors

Beijing, China

Shiwan Zhao