arXiv subjects

Fei Tan

Publications and source records attributed to Fei Tan.

45 records · Page 3Linked to original sources

arXiv2020

HABERTOR: An Efficient and Effective Deep Hatespeech Detector

We present our HABERTOR model for detecting hatespeech in large scale user-generated content. Inspired by the recent success of the BERT model, we propose several modifications to BERT to enhance the performance on the downstream hatespeech classification task. HABERTOR inherits BERT's architecture, but is different in four aspects: (i) it generates its own vocabularies and is pre-trained from the scratch using the largest scale hatespeech dataset; (ii) it consists of Quaternion-based factorized components, resulting in a much smaller number of parameters, faster training and inferencing, as well as less memory usage; (iii) it uses our proposed multi-source ensemble heads with a pooling layer for separate input sources, to further enhance its effectiveness; and (iv) it uses a regularized adversarial training with our proposed fine-grained and adaptive noise magnitude to enhance its robustness. Through experiments on the large-scale real-world hatespeech dataset with 1.4M annotated comments, we show that HABERTOR works better than 15 state-of-the-art hatespeech detection methods, including fine-tuning Language Models. In particular, comparing with BERT, our HABERTOR is 4~5 times faster in the training/inferencing phase, uses less than 1/3 of the memory, and has better performance, even though we pre-train it by using less than 1% of the number of words. Our generalizability analysis shows that HABERTOR transfers well to other unseen hatespeech datasets and is a more efficient and effective alternative to BERT for the hatespeech classification.

cs.CL↗

arXiv2020

DeepVar: An End-to-End Deep Learning Approach for Genomic Variant Recognition in Biomedical Literature

We consider the problem of Named Entity Recognition (NER) on biomedical scientific literature, and more specifically the genomic variants recognition in this work. Significant success has been achieved for NER on canonical tasks in recent years where large data sets are generally available. However, it remains a challenging problem on many domain-specific areas, especially the domains where only small gold annotations can be obtained. In addition, genomic variant entities exhibit diverse linguistic heterogeneity, differing much from those that have been characterized in existing canonical NER tasks. The state-of-the-art machine learning approaches in such tasks heavily rely on arduous feature engineering to characterize those unique patterns. In this work, we present the first successful end-to-end deep learning approach to bridge the gap between generic NER algorithms and low-resource applications through genomic variants recognition. Our proposed model can result in promising performance without any hand-crafted features or post-processing rules. Our extensive experiments and results may shed light on other similar low-resource NER applications.

cs.CL↗

arXiv2018

A Blended Deep Learning Approach for Predicting User Intended Actions

User intended actions are widely seen in many areas. Forecasting these actions and taking proactive measures to optimize business outcome is a crucial step towards sustaining the steady business growth. In this work, we focus on pre- dicting attrition, which is one of typical user intended actions. Conventional attrition predictive modeling strategies suffer a few inherent drawbacks. To overcome these limitations, we propose a novel end-to-end learning scheme to keep track of the evolution of attrition patterns for the predictive modeling. It integrates user activity logs, dynamic and static user profiles based on multi-path learning. It exploits historical user records by establishing a decaying multi-snapshot technique. And finally it employs the precedent user intentions via guiding them to the subsequent learning procedure. As a result, it addresses all disadvantages of conventional methods. We evaluate our methodology on two public data repositories and one private user usage dataset provided by Adobe Creative Cloud. The extensive experiments demonstrate that it can offer the appealing performance in comparison with several existing approaches as rated by different popular metrics. Furthermore, we introduce an advanced interpretation and visualization strategy to effectively characterize the periodicity of user activity logs. It can help to pinpoint important factors that are critical to user attrition and retention and thus suggests actionable improvement targets for business practice. Our work will provide useful insights into the prediction and elucidation of other user intended actions as well.

cs.LG↗

arXiv2014

Link Prediction in Complex Networks: A Mutual Information Perspective

Topological properties of networks are widely applied to study the link-prediction problem recently. Common Neighbors, for example, is a natural yet efficient framework. Many variants of Common Neighbors have been thus proposed to further boost the discriminative resolution of candidate links. In this paper, we reexamine the role of network topology in predicting missing links from the perspective of information theory, and present a practical approach based on the mutual information of network structures. It not only can improve the prediction accuracy substantially, but also experiences reasonable computing complexity.

cs.SI↗

arXiv2014

The robust-yet-fragile nature of interdependent networks

Interdependent networks have been shown to be extremely vulnerable based on the percolation model. Parshani et. al further indicated that the more inter-similar networks are, the more robust they are to random failure. Our understanding of how coupling patterns shape and impact the cascading failures of loads in interdependent networks is limited, but is essential for the design and optimization of the real-world interdependent networked systems. This question, however, is largely unexplored. In this paper, we address this question by investigating the robustness of interdependent ER random graphs and BA scale-free networks under both random failure and intentional attack. It is found that interdependent ER random graphs are robust-yet-fragile under both random failures and intentional attack. Interdependent BA scale-free networks, however, are only robust-yet-fragile under random failure but fragile under intentional attack. These results advance our understanding of the robustness of interdependent networks significantly.

physics.soc-ph↗

arXiv2014

Traffic congestion in interconnected complex networks

Traffic congestion in isolated complex networks has been investigated extensively over the last decade. Coupled network models have recently been developed to facilitate further understanding of real complex systems. Analysis of traffic congestion in coupled complex networks, however, is still relatively unexplored. In this paper, we try to explore the effect of interconnections on traffic congestion in interconnected BA scale-free networks. We find that assortative coupling can alleviate traffic congestion more readily than disassortative and random coupling when the node processing capacity is allocated based on node usage probability. Furthermore, the optimal coupling probability can be found for assortative coupling. However, three types of coupling preferences achieve similar traffic performance if all nodes share the same processing capacity. We analyze interconnected Internet AS-level graphs of South Korea and Japan and obtain similar results. Some practical suggestions are presented to optimize such real-world interconnected networks accordingly.

physics.soc-ph↗

arXiv2007

Two-mode variational Monte Carlo study of Quasiparticle excitations in cuprates

Recent measurements of quasiparticles in hole-doped cuprates reveal highly unusual features: 1) the doping-independent Fermi velocity, 2) two energy scales in the quasiparticle spectral function, and 3) a suppression of the low energy spectral weight near the zone center. The underlying mechanism is under hot debate. We addressed these important issues by a novel two-mode variational Monte Carlo (VMC) study of the t-J model. We obtained results in agreement with the experiments but without invoking extrinsic effects. Besides, we resolved a long standing issue of the sum rule for quasiparticle spectral weights in VMC studies. The electron doped case was also discussed.

cond-mat.supr-con↗

arXiv2006

Theory of high energy features in angle-resolved photo-emission spectra of hole-doped cuprates

The recent angle-resolved photoemission measurements performed up to binding energies of the order of 1eV reveals a very robust feature: the nodal quasi-particle dispersion breaks up around 0.3-0.4eV and reappears around 0.6-0.8eV. The intensity map in the energy-momentum space shows a waterfall like feature between these two energy scales. We argue and numerically demonstrate that these experimental features follow naturally from the strong correlation effects built in the familiar t-J model, and reflect the connection between the fermi level and the lower Hubbard band. The results were obtained by a mean field theory that effectively projects electrons by quantum interference between two bands of fermions instead of binding slave particles.

cond-mat.supr-con↗

arXiv2005

Electronic Raman scattering in superconductors on triangle lattices

In superconductors on triangle lattices, we show that the electronic Raman spectra in the non-resonant channel are independent of the light polarization configuration in the scattering. Therefore the light polarization configuration is no longer a judicious choice to probe the anisotropic nature of the pairing wave function (as in the case of square lattices). However, we show that the detailed energy-shift dependence of the spectrum and the effects of impurities provide useful information that can help identify the possible pairing symmetry experimentally. These results are relevant to the newly discovered Na$_x$CoO$_2\cdot y$H$_2$O superconductors.

cond-mat.supr-con↗