arXiv subjects

Tong Wang

Publications and source records attributed to Tong Wang.

At least 163 records · Page 9Linked to original sources

arXiv2020

Exploring and Predicting Transferability across NLP Tasks

Recent advances in NLP demonstrate the effectiveness of training large-scale language models and transferring them to downstream tasks. Can fine-tuning these models on tasks other than language modeling further improve performance? In this paper, we conduct an extensive study of the transferability between 33 NLP tasks across three broad classes of problems (text classification, question answering, and sequence labeling). Our results show that transfer learning is more beneficial than previously thought, especially when target task data is scarce, and can improve performance even when the source task is small or differs substantially from the target task (e.g., part-of-speech tagging transfers well to the DROP QA dataset). We also develop task embeddings that can be used to predict the most transferable source tasks for a given target task, and we validate their effectiveness in experiments controlled for source and target data size. Overall, our experiments reveal that factors such as source data size, task and domain similarity, and task complexity all play a role in determining transferability.

cs.CL↗

arXiv2020

Monetizing Edge Service in Mobile Internet Ecosystem

In mobile Internet ecosystem, Mobile Users (MUs) purchase wireless data services from Internet Service Provider (ISP) to access to Internet and acquire the interested content services (e.g., online game) from Content Provider (CP). The popularity of intelligent functions (e.g., AI and 3D modeling) increases the computation-intensity of the content services, leading to a growing computation pressure for the MUs' resource-limited devices. To this end, edge computing service is emerging as a promising approach to alleviate the MUs' computation pressure while keeping their quality-of-service, via offloading some computation tasks of MUs to edge (computing) servers deployed at the local network edge. Thus, Edge Service Provider (ESP), who deploys the edge servers and offers the edge computing service, becomes an upcoming new stakeholder in the ecosystem. In this work, we study the economic interactions of MUs, ISP, CP, and ESP in the new ecosystem with edge computing service, where MUs can acquire the computation-intensive content services (offered by CP) and offload some computation tasks, together with the necessary raw input data, to edge servers (deployed by ESP) through ISP. We first study the MU's Joint Content Acquisition and Task Offloading (J-CATO) problem, which aims to maximize his long-term payoff. We derive the off-line solution with crucial insights, based on which we design an online strategy with provable performance. Then, we study the ESP's edge service monetization problem. We propose a pricing policy that can achieve a constant fraction of the ex-post optimal revenue with an extra constant loss for the ESP. Numerical results show that the edge computing service can stimulate the MUs' content acquisition and improve the payoffs of MUs, ISP, and CP.

cs.GT↗

arXiv2020

One Size Does Not Fit All: Generating and Evaluating Variable Number of Keyphrases

Different texts shall by nature correspond to different number of keyphrases. This desideratum is largely missing from existing neural keyphrase generation models. In this study, we address this problem from both modeling and evaluation perspectives. We first propose a recurrent generative model that generates multiple keyphrases as delimiter-separated sequences. Generation diversity is further enhanced with two novel techniques by manipulating decoder hidden states. In contrast to previous approaches, our model is capable of generating diverse keyphrases and controlling number of outputs. We further propose two evaluation metrics tailored towards the variable-number generation. We also introduce a new dataset StackEx that expands beyond the only existing genre (i.e., academic writing) in keyphrase generation tasks. With both previous and new evaluation metrics, our model outperforms strong baselines on all datasets.

cs.CL↗

arXiv2020

Efficient Estimation of Mixture Cure Frailty Model for Clustered Current Status Data

Current status data abounds in the field of epidemiology and public health, where the only observable data for a subject is the random inspection time and the event status at inspection. Motivated by such a current status data from a periodontal study where data are inherently clustered, we propose a unified methodology to analyze such complex data. We allow the time-to-event to follow the semiparametric GOR model with a cure fraction, and develop a unified estimation scheme powered by the EM algorithm. The within-subject correlation is accounted for by a random (frailty) effect, and the non-parametric component of the GOR model is approximated via penalized splines, with a set of knot points that increases with the sample size. Proposed methodology is accompanied by a rigorous asymptotic theory, and the related semiparametric efficiency. The finite sample performance of our model parameters are assessed via simulation studies. Furthermore, the proposed methodology is illustrated via application to the oral health data, accompanied by diagnostic checks to identify influential observations. An easy to use R package CRFCSD is also available for implementation.

stat.ME↗

arXiv2020

Crowd-MECS: A Novel Crowdsourcing Framework for Mobile Edge Caching and Sharing

Crowdsourced mobile edge caching and sharing (Crowd-MECS) is emerging as a promising content delivery paradigm by employing a large crowd of existing edge devices (EDs) to cache and share popular contents. The successful technology adoption of Crowd-MECS relies on a comprehensive understanding of the complicated economic interactions and strategic decision-making of different stakeholders. In this paper, we focus on studying the economic and strategic interactions between one content provider (CP) and a large crowd of EDs, where the EDs can decide whether to cache and share contents for the CP, and the CP can decide to share a certain revenue with EDs as the incentive of caching and sharing contents. We formulate such an interaction as a two-stage Stackelberg game. In Stage I, the CP aims to maximize its own profit by deciding the ratio of revenue shared with EDs. In Stage II, EDs aim to maximize their own payoffs by choosing to be agents who cache and share contents, and meanwhile gain a certain revenue from the CP, or requesters who do not cache but request contents in the on-demand fashion. We first analyze the EDs' best responses and prove the existence and uniqueness of the equilibrium in Stage II by using the non-atomic game theory. Then, we identify the piece-wise structure and the unimodal feature of the CP's profit function, based on which we design a tailored low-complexity one-dimensional search algorithm to achieve the optimal revenue sharing ratio for the CP in Stage I. Simulation results show that both the CP's profit and the EDs' total welfare can be improved significantly (e.g., by 120% and 50%, respectively) by using the proposed Crowd-MECS, comparing with the Non-MEC system where the CP serves all EDs directly.

cs.GT↗

arXiv2020

Optimizing Traffic Lights with Multi-agent Deep Reinforcement Learning and V2X communication

We consider a system to optimize duration of traffic signals using multi-agent deep reinforcement learning and Vehicle-to-Everything (V2X) communication. This system aims at analyzing independent and shared rewards for multi-agents to control duration of traffic lights. A learning agent traffic light gets information along its lanes within a circular V2X coverage. The duration cycles of traffic light are modeled as Markov decision Processes. We investigate four variations of reward functions. The first two are unshared-rewards: based on waiting number, and waiting time of vehicles between two cycles of traffic light. The third and fourth functions are: shared-rewards based on waiting cars, and waiting time for all agents. Each agent has a memory for optimization through target network and prioritized experience replay. We evaluate multi-agents through the Simulation of Urban MObility (SUMO) simulator. The results prove effectiveness of the proposed system to optimize traffic signals and reduce average waiting cars to 41.5 % as compared to the traditional periodic traffic control system.

cs.AI↗

arXiv2020

Modeling microbial cross-feeding at intermediate scale portrays community dynamics and species coexistence

Social interaction between microbes can be described at many levels of details, ranging from the biochemistry of cell-cell interactions to the ecological dynamics of populations. Choosing the best level to model microbial communities without losing generality remains a challenge. Here we propose to model cross-feeding interactions at an intermediate level between genome-scale metabolic models of individual species and consumer-resource models of ecosystems, which is suitable to empirical data. We applied our method to three published examples of multi-strain Escherichia coli communities with increasing complexity consisting of uni-, bi-, and multi-directional cross-feeding of either substitutable metabolic byproducts or essential nutrients. The intermediate-scale model accurately described empirical data and could quantify exchange rates elusive by other means, such as the byproduct secretions, even for a complex community of 14 amino acid auxotrophs. We used the three models to study each community's limits of robustness to perturbations such as variations in resource supply, antibiotic treatments and invasion by other "cheaters" species. Our analysis provides a foundation to quantify cross-feeding interactions from experimental data, and highlights the importance of metabolic exchanges in the dynamics and stability of microbial communities.

q-bio.PE↗

arXiv2020

Interpretable Companions for Black-Box Models

We present an interpretable companion model for any pre-trained black-box classifiers. The idea is that for any input, a user can decide to either receive a prediction from the black-box model, with high accuracy but no explanations, or employ a companion rule to obtain an interpretable prediction with slightly lower accuracy. The companion model is trained from data and the predictions of the black-box model, with the objective combining area under the transparency--accuracy curve and model complexity. Our model provides flexible choices for practitioners who face the dilemma of choosing between always using interpretable models and always using black-box models for a predictive task, so users can, for any given input, take a step back to resort to an interpretable prediction if they find the predictive performance satisfying, or stick to the black-box model if the rules are unsatisfying. To show the value of companion models, we design a human evaluation on more than a hundred people to investigate the tolerable accuracy loss to gain interpretability for humans.

stat.ML↗

arXiv2020

Anomalous Phonon-mode Dependence in Polarized Raman Spectroscopy of Topological Weyl Semimetal TaP

Topological Weyl semimetals (WSMs) have attracted widespread interests due to the chiral Weyl fermions and surface Fermi arcs that enable unique optical and transport phenomena. In this work, we present angle-resolved Raman spectroscopy of TaP, a prototypical noncentrosymmetric WSM, for five excitation wavelengths ranging from 364 to 785 nm. The Raman active modes, $A_1$, $B_1^1$, and $B_1^2$ modes, exhibit two main unique features beyond the conventional Raman theory. First, the relative intensities of Raman active modes change as a function of the excitation wavelength. Second, angle-resolved polarized Raman spectra show systematic deviation from the Raman tensor theory. In particular, the $B_1^1$ mode is absent for 633 nm excitation, whereas the $B_1^2$ mode shows an unusual two-fold symmetry instead of a four-fold symmetry for 488, 532, and 633 nm excitations. These unconventional phenomena are attributed to the interference effect in the Raman process owing to the existence of multiple carrier pockets with almost the same energy but different symmetries.

cond-mat.mtrl-sci↗

arXiv2019

Metalearned Neural Memory

We augment recurrent neural networks with an external memory mechanism that builds upon recent progress in metalearning. We conceptualize this memory as a rapidly adaptable function that we parameterize as a deep neural network. Reading from the neural memory function amounts to pushing an input (the key vector) through the function to produce an output (the value vector). Writing to memory means changing the function; specifically, updating the parameters of the neural network to encode desired information. We leverage training and algorithmic techniques from metalearning to update the neural memory function in one shot. The proposed memory-augmented model achieves strong performance on a variety of learning problems, from supervised question answering to reinforcement learning.

cs.NE↗

arXiv2019

Early Predictions for Medical Crowdfunding: A Deep Learning Approach Using Diverse Inputs

Medical crowdfunding is a popular channel for people needing financial help paying medical bills to collect donations from large numbers of people. However, large heterogeneity exists in donations across cases, and fundraisers face significant uncertainty in whether their crowdfunding campaigns can meet fundraising goals. Therefore, it is important to provide early warnings for fundraisers if such a channel will eventually fail. In this study, we aim to develop novel algorithms to provide accurate and timely predictions of fundraising performance, to better inform fundraisers. In particular, we propose a new approach to combine time-series features and time-invariant features in the deep learning model, to process diverse sources of input data. Compared with baseline models, our model achieves better accuracy and requires a shorter observation window of the time-varying features from the campaign launch to provide robust predictions with high confidence. To extract interpretable insights, we further conduct a multivariate time-series clustering analysis and identify four typical temporal donation patterns. This demonstrates the heterogeneity in the features and how they relate to the fundraising outcome. The prediction model and the interpretable insights can be applied to assist fundraisers with better promoting their fundraising campaigns and can potentially help crowdfunding platforms to provide more timely feedback to all fundraisers. Our proposed framework is also generalizable to other fields where diverse structured and unstructured data are valuable for predictions.

cs.LG↗

arXiv2019

Model-Agnostic Linear Competitors -- When Interpretable Models Compete and Collaborate with Black-Box Models

Driven by an increasing need for model interpretability, interpretable models have become strong competitors for black-box models in many real applications. In this paper, we propose a novel type of model where interpretable models compete and collaborate with black-box models. We present the Model-Agnostic Linear Competitors (MALC) for partially interpretable classification. MALC is a hybrid model that uses linear models to locally substitute any black-box model, capturing subspaces that are most likely to be in a class while leaving the rest of the data to the black-box. MALC brings together the interpretable power of linear models and good predictive performance of a black-box model. We formulate the training of a MALC model as a convex optimization. The predictive accuracy and transparency (defined as the percentage of data captured by the linear models) balance through a carefully designed objective function and the optimization problem is solved with the accelerated proximal gradient method. Experiments show that MALC can effectively trade prediction accuracy for transparency and provide an efficient frontier that spans the entire spectrum of transparency.

cs.LG↗

arXiv2019

Reduction of radiative lifetime and slow-timescale spectral diffusion in InGaN polarized single-photon sources

Non-polar (11-20) a-plane quantum dots (QDs) are strong candidates for both > 200 K on-chip ultrafast polarized single-photon generation and the investigation of high temperature semiconductor QD photophysics. In this work, we report progress in the growth of a-plane InGaN QDs with a quasi-two-temperature method, which produces smooth epilayers and significantly reduced carrier trapping sites in the proximity of the QDs. Optical characterization has confirmed the ability of such QDs to emit polarized single photons and we have recorded a ~ 45% shorter average radiative lifetime and 65% reduction in the slow-timescale spectral diffusion compared to previous QDs. This growth method is an important development of the non-polar a-plane InGaN platform, opening up more possibilities in single-photon, lasing, and fundamental investigations.

cond-mat.mes-hall↗

arXiv2019

Evidence for a multi-level trophic organization of the human gut microbiome

The human gut microbiome is a complex ecosystem, in which hundreds of microbial species and metabolites coexist, in part due to an extensive network of cross-feeding interactions. However, both the large-scale trophic organization of this ecosystem, and its effects on the underlying metabolic flow, remain unexplored. Here, using a simplified model, we provide quantitative support for a multi-level trophic organization of the human gut microbiome, where microbes consume and secrete metabolites in multiple iterative steps. Using a manually-curated set of metabolic interactions between microbes, our model suggests about four trophic levels, each characterized by a high level-to-level metabolic transfer of byproducts. It also quantitatively predicts the typical metabolic environment of the gut (fecal metabolome) in approximate agreement with the real data. To understand the consequences of this trophic organization, we quantify the metabolic flow and biomass distribution, and explore patterns of microbial and metabolic diversity in different levels. The hierarchical trophic organization suggested by our model can help mechanistically establish causal links between the abundances of microbes and metabolites in the human gut.

q-bio.PE↗

arXiv2019

Improved fragment-based movement with LRFragLib for all-atom Ab initio protein folding

Fragment-based assembly has been widely used in Ab initio protein folding simulation which can effectively reduce the conformational space and thus accelerate sampling. The efficiency of fragment-based movement as well as the quality of fragment library determine whether the folding process can lead the free energy landscape to the global minimum and help the protein to reach near-native folded state. We designed an improved fragment-based movement, "fragmove", which substituted multiple backbone dihedral angles in every simulation step. This movement strategy was derived from the fragment library generated by LRFragLib, an effective fragment detection algorithm using logistic regression model. We show in replica exchange Monte Carlo (REMC) simulation that "fragmove", when compared with a set of existing movements in REMC, shows significant improved ability at increasing secondary and tertiary predicted model accuracy by 11.24% and 17.98%, respectively and reaching energy minima decreased by 5.72%. Our results demonstrate that this improved movement is more powerful to guide proteins faster to low energy regions of conformational space and promote folding efficiency and predicted model accuracy.

q-bio.QM↗

arXiv2019

Gaining Free or Low-Cost Transparency with Interpretable Partial Substitute

This work addresses the situation where a black-box model with good predictive performance is chosen over its interpretable competitors, and we show interpretability is still achievable in this case. Our solution is to find an interpretable substitute on a subset of data where the black-box model is overkill or nearly overkill while leaving the rest to the black-box. This transparency is obtained at minimal cost or no cost of the predictive performance. Under this framework, we develop a Hybrid Rule Sets (HyRS) model that uses decision rules to capture the subspace of data where the rules are as accurate or almost as accurate as the black-box provided. To train a HyRS, we devise an efficient search algorithm that iteratively finds the optimal model and exploits theoretically grounded strategies to reduce computation. Our framework is agnostic to the black-box during training. Experiments on structured and text data show that HyRS obtains an effective trade-off between transparency and interpretability.

cs.LG↗

arXiv2019

Hybrid Predictive Model: When an Interpretable Model Collaborates with a Black-box Model

Interpretable machine learning has become a strong competitor for traditional black-box models. However, the possible loss of the predictive performance for gaining interpretability is often inevitable, putting practitioners in a dilemma of choosing between high accuracy (black-box models) and interpretability (interpretable models). In this work, we propose a novel framework for building a Hybrid Predictive Model (HPM) that integrates an interpretable model with any black-box model to combine their strengths. The interpretable model substitutes the black-box model on a subset of data where the black-box is overkill or nearly overkill, gaining transparency at no or low cost of the predictive accuracy. We design a principled objective function that considers predictive accuracy, model interpretability, and model transparency (defined as the percentage of data processed by the interpretable substitute.) Under this framework, we propose two hybrid models, one substituting with association rules and the other with linear models, and we design customized training algorithms for both models. We test the hybrid models on structured data and text data where interpretable models collaborate with various state-of-the-art black-box models. Results show that hybrid models obtain an efficient trade-off between transparency and predictive performance, characterized by our proposed efficient frontiers.

cs.LG↗

arXiv2018

Next Hit Predictor - Self-exciting Risk Modeling for Predicting Next Locations of Serial Crimes

Our goal is to predict the location of the next crime in a crime series, based on the identified previous offenses in the series. We build a predictive model called Next Hit Predictor (NHP) that finds the most likely location of the next serial crime via a carefully designed risk model. The risk model follows the paradigm of a self-exciting point process which consists of a background crime risk and triggered risks stimulated by previous offenses in the series. Thus, NHP creates a risk map for a crime series at hand. To train the risk model, we formulate a convex learning objective that considers pairwise rankings of locations and use stochastic gradient descent to learn the optimal parameters. Next Hit Predictor incorporates both spatial-temporal features and geographical characteristics of prior crime locations in the series. Next Hit Predictor has demonstrated promising results on decades' worth of serial crime data collected by the Crime Analysis Unit of the Cambridge Police Department in Massachusetts, USA.

stat.AP↗