Nec Corporation

Nec C Orporation

Bo Zong

Haifeng Chen

Wei Cheng

LuAn Tang

Jingchao Ni

Hui Joyce Zhang

Guofei Jiang

Cristian Lumezanu

Jianwu Xu

Nipun Arora

Xuchao Zhang

Yanchi Liu

Daeki Cho

Zhengzhang Chen

Biplob Kumar Debnath

Wenchao Yu

Mingda Li

Zhichun Li

Kenji Yoshihira

Dongjin Song

Qi Song

Hengtong Zhang

Cheng Zheng

Denghui Zhang

Lichen Wang

Hancheng Ge

Youfu Li

Xin Dong

Christian Lumezanu

Junxiang Wang

Zhen Wang

A method trains a recursive reasoning unit (RRU). The method receives a graph for a set of words and a matrix for a different set of words. The graph maps each word in the set of words to a node with node label and indicates a relation between adjacent nodes by an edge with edge label. The matrix indicates word co-occurrence frequency of the different set of words. The method discovers, by the RRU, reasoning paths from the graph for word pairs by mapping word pairs from the set of words into a source word and a destination word and finding the reasoning paths therebetween. The method predicts word co-occurrence frequency using the reasoning paths. The method updates, responsive to the word co-occurrence frequency, model parameters of the RRU until a difference between a predicted and true word occurrence are less than a threshold amount to provide a trained RRU.

Learning word representations via commonsense reasoning

Systems and methods are provided for Cross-lingual Transfer Interpretation (CTI). The method includes receiving text corpus data including premise-hypothesis pairs with a relationship label in a source language, and conducting a source to target language translation. The method further includes performing a feature importance extraction, where an integrated gradient is applied to assign an importance score to each input feature, and performing a cross-lingual feature alignment, where tokens in the source language are aligned with tokens in the target language for both the premise and the hypothesis based on semantic similarity. The method further includes performing a qualitative analysis, where the importance score of each token can be compared between the source language and the target language according to a feature alignment result.

Interpreting cross-lingual models for natural language inference

A computer-implemented method is provided for cross-lingual transfer. The method includes randomly masking a source corpus and a target corpus to obtain a masked source corpus and a masked target corpus. The method further includes tokenizing, by pretrained Natural Language Processing (NLP) models, the masked source corpus and the masked target corpus to obtain source tokens and target tokens. The method also includes transforming the source tokens and the target tokens into a source dependency parsing tree and a target dependency parsing tree. The method additionally includes inputting the source dependency parsing tree and the target dependency parsing tree into a graph encoder pretrained on a translation language modeling task to extract common language information for transfer. The method further includes fine-tuning the graph encoder and a down-stream network for a specific NLP down-stream task.

Cross-lingual zero-shot transfer via semantic and synthetic representation learning

Methods and systems for natural language processing include generating an encoder that includes a global part and a local part, where the global part encodes multi-hop relations between words in an input and where the local part encodes one-hop relations between words in the input. The encoder is trained to form a graph that represents tokens of an input text as nodes and that represents relations between the tokens as edges between the nodes.

Graph-based cross-lingual zero-shot transfer

Methods and systems for allocating network resources responsive to network traffic include modeling spatial correlations between fine spatial granularity traffic and coarse spatial granularity traffic for different sites and regions to determine spatial feature vectors for one or more sites in a network. Temporal correlations at a fine spatial granularity are modeled across multiple temporal scales, based on the spatial feature vectors. Temporal correlations at a coarse spatial granularity are modeled across multiple temporal scales, based on the spatial feature vectors. A traffic flow prediction is determined for the one or more sites in the network, based on the temporal correlations at the fine spatial granularity and the temporal correlations at the coarse spatial granularity. Network resources are provisioned at the one or more sites in accordance with the traffic flow prediction.

Multi-scale multi-granularity spatial-temporal traffic volume prediction

A method for employing a temporal context-aware question routing model (TCQR) in multiple granularities of temporal dynamics in community-based question answering (CQA) systems is presented. The method includes encoding answerers into temporal context-aware representations based on semantic and temporal information of questions, measuring answerers expertise in one or more of the questions as a coherence between the temporal context-aware representations of the answerers and encodings of the questions, modeling the temporal dynamics of answering behaviors of the answerers in different levels of time granularities by using multi-shift and multi-resolution extensions, and outputting answers of select answerers to a visualization device.

Temporal context-aware representation learning for question routing

A method for employing a unified semi-supervised deep learning (DL) framework for turbulence forecasting is presented. The method includes extracting historical and forecasted weather features of a spatial region, calculating turbulence indexes to fill feature cubes, each feature cube representing a grid-based 3D region, and building an encoder-decoder framework based on convolutional long short-term memory (ConvLSTM) to model spatio-temporal correlations or patterns causing turbulence. The method further includes employing a dual label guessing component to dynamically integrate complementary signals from a turbulence forecasting network and a turbulence detection network to generate pseudo-labels, reweighing the generated pseudo-labels by a heuristic label quality detector based on KL-Divergence, applying a hybrid loss function to predict turbulence conditions, and generating a turbulence dataset including the predicted turbulence conditions.

Semi-supervised deep model for turbulence forecasting

A method for employing a differentiable ranking based graph sparsification (DRGS) network to use supervision signals from downstream tasks to guide graph sparsification is presented. The method includes, in a training phase, generating node representations by neighborhood aggregation operators, generating sparsified subgraphs by top-k neighbor sampling from a learned neighborhood ranking distribution, feeding the sparsified subgraphs to a task, generating a prediction, and collecting a prediction error to update parameters in the generating and feeding steps to minimize an error, and, in a testing phase, generating node representations by neighborhood aggregation operators related to testing data, generating sparsified subgraphs by top-k neighbor sampling from a learned neighborhood ranking distribution related to the testing data, feeding the sparsified subgraphs related to the testing data to a task, and outputting prediction results to a visualization device.

Deep graph de-noise by differentiable ranking

A computer-implemented method executed by at least one processor for software bug localization is presented. The method includes constructing a bug localization graph to capture relationships between bug tickets and relevant source code files from historical change-sets and an underlying source code repository, leveraging natural processing language tools to evaluate semantic similarity between a new bug ticket and a historical ticket, in response to the evaluated semantic similarity, for the new bug ticket, adding links between the new bug ticket a set of similar historical tickets, incorporating the new bug ticket in the bug localization graph, and developing a mathematical graph expression to determine a closeness relationship between the relevant source code files and the new bug ticket.

Graph-based method for inductive bug localization

A method for employing a supervised graph sparsification (SGS) network to use feedback from subsequent graph learning tasks to guide graph sparsification is presented. The method includes, in a training phase, generating sparsified subgraphs by edge sampling from input training graphs following a learned distribution, feeding the sparsified subgraphs to a prediction/classification component, collecting a predication/classification error, and updating parameters of the learned distribution based on a gradient derived from the predication/classification error. The method further includes, in a testing phase, generating sparsified subgraphs by edge sampling from input testing graphs following the learned distribution, feeding the sparsified subgraphs to the prediction/classification component, and outputting prediction/classification results to a visualization device.

Method for supervised graph sparsification

Methods and systems for detecting and responding to anomalous nodes in a network include inferring temporal factors, using a computer-implemented neural network, that represent changes in a network graph across time steps, with a temporal factor for each time step depending on a temporal factor for a previous time step. An invariant factor is inferred that represents information about the network graph that does not change across the time steps. The temporal factors and the invariant factor are combined into a combined temporal-invariant representation. It is determined that an unlabeled node is anomalous, based on the combined temporal-invariant representation. A security action is performed responsive to the determination that unlabeled node is anomalous.

Node classification in dynamic networks using graph factorization

Systems and methods for predicting system device failure are provided. The method includes representing device failure related data associated with the devices from a predetermined domain by temporal graphs for each of the devices. The method also includes extracting vector representations based on temporal graph features from the temporal graphs that capture both temporal and structural correlation in the device failure related data. The method further includes predicting, based on the vector representations and device failure related metrics in the predetermined domain, one or more of the devices that is expected to fail within a predetermined time.

Performance prediction from communication data

Methods and systems for detecting abnormal application behavior include determining a vector representation of a first syscall graph that is generated by a first application, the vector representation including a representation of a distribution of subgraphs of the first syscall graph. The vector representation of the first syscall graph is compared to one or more second syscall graphs that are generated by respective second applications to determine respective similarity scores. It is determined that the first application is behaving abnormally based on the similarity scores, and a security action is performed responsive to the determination that the first application is behaving abnormally.

Unsupervised graph similarity learning based on stochastic subgraph sampling

A method for implementing a recommendation system using an asymmetrically hierarchical network includes, for a user and an item corresponding to a user-item pair, aggregating, using asymmetrically designed sentence aggregators, respective ones of a set of item sentence embeddings and a set of user sentence embeddings to generate a set of item review embeddings based on first item attention weights and a set of user review embeddings based on first user attention weights, respectively, aggregating, using asymmetrically designed review aggregators, respective ones of the set of item review embeddings and the set of user review embeddings to generate an item embedding based on a second item attention weights and a user embedding based on second user attention weights, respectively, and predicting a rating of the user-item pair based on the item embedding and the user embedding.

Asymmetrically hierarchical networks with attentive interactions for interpretable review-based recommendation

Systems and methods for implementing dynamic graph analysis (DGA) to detect anomalous network traffic are provided. The method includes processing communications and profile data associated with multiple devices to determine dynamic graphs. The method includes generating features to model temporal behaviors of network traffic generated by the multiple devices based on the dynamic graphs. The method also includes formulating a list of prediction results for sources of the anomalous network traffic from the multiple devices based on the temporal behaviors.

Growing community of inventors

Plainsboro, NJ, United States of America

Bo Zong