Microblogs have become popular media for news propagation in recent years. Meanwhile, numerous rumors and fake also bloom spread wildly on the open social platforms. Without verification, they could seriously jeopardize credibility of microblogs. We observe that an increasing number users are using images videos to post addition texts. Tweets or microblogs commonly composed text, image context. In this paper, we propose a novel Recurrent Neural Network with attention mechanism (att-RNN) fuse...

Microblogs have become one of the most popular platforms for news sharing. However, due to its openness and lack supervision, rumors could also be easily posted propagated on social networks, which cause huge panic threat during propagation. In this paper, we detect by leveraging hierarchical representations at different levels contexts. Specifically, propose a novel neural network combined with information (HSA-BLSTM). We first build bidirectional long short-term memory model representation...

Flexible strain sensors have attracted a great amount of attention for promising applications in next-generation artificially intelligent devices. However, it is difficult conventional planar to meet the requirements miniature size and light weight flexible electronics. Herein, highly sensitive stretchable fiber sensor with millimeter diameter was innovatively fabricated by capillary tube method integrate silver nanowires (AgNWs) polyurethane (PU) fibers. Scanning electron microscopy results...

The limited availability of annotated data often hinders real-world applications machine learning. To efficiently learn from small quantities multimodal data, we leverage the linguistic knowledge a large pre-trained language model (PLM) and quickly adapt it to new domains image captioning. effectively utilize pretrained model, is critical balance visual input prior pretraining. We propose VisualGPT, which employs novel self-resurrecting encoder-decoder attention mechanism PLM with amount...

An accurate abstractive summary of a document should contain all its salient information and be logically entailed by the input document. We improve these important aspects summarization via multi-task learning with auxiliary tasks question generation entailment generation, where former teaches model how to look for questioning-worthy details, latter rewrite which is directed-logical subset also propose novel architectures high-level (semantic) layer-specific sharing across multiple encoder...

This paper designs and evaluates a practical algorithm, called recursive projected compressive sensing (Prac-ReProCS), for recovering time sequence of sparse vectors $S_t$ dense $L_t$ from their sum, $M_t:= S_t + L_t$, when any subsequence the $L_t$'s lies in slowly changing low-dimensional subspace. A key application where this problem occurs is video layering goal to separate into background foreground that consists one or more moving regions/objects. Prac-ReProCS modification its...

A butterfly-based direct combined-field integral-equation (CFIE) solver for analyzing scattering from electrically large, perfect conducting objects is presented. The proposed leverages the butterfly scheme to compress blocks of hierarchical LU-factorized discretized CFIE operator and uses randomized reconstruction schemes expedite factorization. memory requirement computational cost butterfly-CFIE scale as O(N log <sup xmlns:mml="http://www.w3.org/1998/Math/MathML"...

Domain adaptation performance of a learning algorithm on target domain is function its source error and divergence measure between the data distribution these two domains. We present study various distance-based measures in context NLP tasks, that characterize dissimilarity domains based sample estimates. first conduct analysis experiments to show which distance can best differentiate samples from same versus different domains, are correlated with empirical results. Next, we develop...

The ever-increasing amount of multimedia content on modern social media platforms are valuable in many applications. While the openness and convenience features also foster rumors online. Without verification, these would reach thousands users immediately cause serious damages. Many efforts have been taken to defeat online automatically by mining rich provided open network with machine learning techniques. Most rumor detection methods can be categorized three paradigms: hand-crafted based...

Emojis, as a new way of conveying nonverbal cues, are widely adopted in computer-mediated communications. In this paper, first from message sender perspective, we focus on people's motives using four types emojis — positive, neutral, negative, and non-facial. We compare the willingness levels these emoji for seven typical intentions that people usually apply cues communication. The results extensive statistical hypothesis tests not only report popularities intentions, but also uncover subtle...

Influence functions approximate the “influences” of training data-points for test predictions and have a wide variety applications. Despite popularity, their computational cost does not scale well with model data size. We present FastIF, set simple modifications to influence that significantly improves run-time. use k-Nearest Neighbors (kNN) narrow search space down subset good candidate points, identify configurations best balance speed-quality trade-off in estimating inverse Hessian-vector...

The features of social rating in this article are disclosed the context its influence on politics a modern state. Social is system control over activity citizens, which assessed by several parame-ters. These parameters analyzed article. advantages (reduction crime rate and number offenses, economic growth, reduction inequality) listed, as well weaknesses (dif-ficulties combining with concept human rights, formal performance good deeds, circumvention system). experience introducing territory...

An efficient electromagnetic interference (EMI) shielding paper with excellent water repellency and mechanical flexibility has been developed, by assembling silver nanowires (AgNWs) hydrophobic inorganic ceramic on the cellulose paper, via a facile dip-coating preparation. Scanning electron microscope (SEM) observations confirmed that AgNWs were interconnected densely coated both sides of fiber, which endows as-prepared high conductivity (33.69 S/cm in-plane direction) at low AgNW area...

With increasing deployment of machine learning systems in various real-world tasks, there is a greater need for accurate quantification predictive uncertainty. While the common goal uncertainty (UQ) to approximate true distribution target data, many works UQ tend be disjoint evaluation metrics utilized, and disparate implementations each metric lead numerical results that are not directly comparable across different works. To address this, we introduce Uncertainty Toolbox, an open-source...

The multilevel matrix decomposition algorithm (MLMDA) is shown to permit effective compression of inverse integral operators pertinent the analysis scattering from electrically large structures. Observed ratios exceed those realized by low-rank (LR) methods, leading substantial memory savings and a faster application operator, suggesting new for schemes traditionally used compressing forward operators.

Emojis, as a new way of conveying nonverbal cues, are widely adopted in computer-mediated communications. In this paper, first from message sender perspective, we focus on people's motives using four types emojis -- positive, neutral, negative, and non-facial. We compare the willingness levels these emoji for seven typical intentions that people usually apply cues communication. The results extensive statistical hypothesis tests not only report popularities intentions, but also uncover...

ive summarization, the task of rewriting and compressing a document into short summary, has achieved considerable success with neural sequence-to-sequence models. However, these models can still benefit from stronger natural language inference skills, since correct summary is logically entailed by input document, i.e., it should not contain any contradictory or unrelated information. We incorporate such knowledge an abstractive summarization model via multi-task learning, where we share its...

A butterfly-based fast direct integral equation solver for analyzing high-frequency scattering from two-dimensional objects is presented. The leverages a randomized butterfly scheme to compress blocks corresponding near- and far-field interactions in the discretized forward inverse electric field operators. observed memory requirements computational cost of proposed scale as O(Nlog^2N) O(N^1.5 logN), respectively. applied analysis electrically large spanning over ten thousand wavelengths...

The limited spatial sampling rates of conventional Shack–Hartmann wavefront sensors (SHWFSs) make them unable to sense higher-order distortion. In this study, by etching a known phase on each microlens modulate sub-wavefront, we propose higher-resolution reconstruction method that employs modified modal Zernike algorithm, in which the matrix contains quadratic information is extracted using neural network. We validate through simulations, and results show once network has been trained, for...

In this paper, a novel technique is proposed to solve the electromagnetic scattering by large finite arrays combining tangential equivalence principle algorithm (T-EPA) with multilevel fast multipole (MLFMA).The (EPA) kind of domain decomposition scheme for and radiation problems based on integral equation (IE).For array same elements, only one matrix needs be constructed stored.T-EPA has better accuracy than original EPA.But calculation impedance in T-EPA still time consuming.MLFMA speed up...

Han Guo, Ramakanth Pasunuru, Mohit Bansal. Proceedings of the 2019 Conference North American Chapter Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers). 2019.