arXiv subjects

Ying Fan

Publications and source records attributed to Ying Fan.

At least 55 records · Page 3Linked to original sources

arXiv2018

Frustration of signed networks: How does it affect the thermodynamic properties of a system?

Signed networks with positive and negative interaction are widely observed in the real systems. The negative links would induce frustration, then affect global properties of the system. Based on previous studies, frustration of signed networks is investigated and quantified. Frustrations of $\pm J$ (Edwards-Anderson) Ising model with a concentration $p$ of negative bonds, constructed on different networks, such as triangular lattice, square lattice and random regular networks (RRN) with connectivity $k=6$ are estimated by theoretical and numerical approaches. Based on the quantitative measurement of frustration, its effects on phase transitions characterized by order parameter $q_{EA}$ are studied. The relationship of critical temperature $T_c$ with the quantified frustration $μ$ is given by mean-field theory. It shows that $T_c$ decreases linearly with frustration $μ$ . The theory is checked by numerical estimations, such as the Metropolis algorithm and Replica Symmetric Population Dynamics Algorithm. The numerical estimates are consistent well with the mean-field prediction.

cond-mat.stat-mech↗

arXiv2018

Deep Interest Network for Click-Through Rate Prediction

Click-through rate prediction is an essential task in industrial applications, such as online advertising. Recently deep learning based models have been proposed, which follow a similar Embedding\&MLP paradigm. In these methods large scale sparse input features are first mapped into low dimensional embedding vectors, and then transformed into fixed-length vectors in a group-wise manner, finally concatenated together to fed into a multilayer perceptron (MLP) to learn the nonlinear relations among features. In this way, user features are compressed into a fixed-length representation vector, in regardless of what candidate ads are. The use of fixed-length vector will be a bottleneck, which brings difficulty for Embedding\&MLP methods to capture user's diverse interests effectively from rich historical behaviors. In this paper, we propose a novel model: Deep Interest Network (DIN) which tackles this challenge by designing a local activation unit to adaptively learn the representation of user interests from historical behaviors with respect to a certain ad. This representation vector varies over different ads, improving the expressive ability of model greatly. Besides, we develop two techniques: mini-batch aware regularization and data adaptive activation function which can help training industrial deep networks with hundreds of millions of parameters. Experiments on two public datasets as well as an Alibaba real production dataset with over 2 billion samples demonstrate the effectiveness of proposed approaches, which achieve superior performance compared with state-of-the-art methods. DIN now has been successfully deployed in the online display advertising system in Alibaba, serving the main traffic.

stat.ML↗

arXiv2018

Rocket Launching: A Universal and Efficient Framework for Training Well-performing Light Net

Models applied on real time response task, like click-through rate (CTR) prediction model, require high accuracy and rigorous response time. Therefore, top-performing deep models of high depth and complexity are not well suited for these applications with the limitations on the inference time. In order to further improve the neural networks' performance given the time and computational limitations, we propose an approach that exploits a cumbersome net to help train the lightweight net for prediction. We dub the whole process rocket launching, where the cumbersome booster net is used to guide the learning of the target light net throughout the whole training process. We analyze different loss functions aiming at pushing the light net to behave similarly to the booster net, and adopt the loss with best performance in our experiments. We use one technique called gradient block to improve the performance of the light net and booster net further. Experiments on benchmark datasets and real-life industrial advertisement data present that our light model can get performance only previously achievable with more complex models.

stat.ML↗

arXiv2014

Universal Predictability of Mobility Patterns in Cities

Despite the long history of modelling human mobility, we continue to lack a highly accurate approach with low data requirements for predicting mobility patterns in cities. Here, we present a population-weighted opportunities model without any adjustable parameters to capture the underlying driving force accounting for human mobility patterns at the city scale. We use various mobility data collected from a number of cities with different characteristics to demonstrate the predictive power of our model. We find that insofar as the spatial distribution of population is available, our model offers universal prediction of mobility patterns in good agreement with real observations, including distance distribution, destination travel constraints and flux. In contrast, the models that succeed in modelling mobility patterns in countries are not applicable in cities, which suggests that there is a diversity of human mobility at different spatial scales. Our model has potential applications in many fields relevant to mobility behaviour in cities, without relying on previous mobility measurements.

physics.soc-ph↗

arXiv2014

Reconstructing propagation networks with natural diversity and identifying hidden sources

Our ability to uncover complex network structure and dynamics from data is fundamental to understanding and controlling collective dynamics in complex systems. Despite recent progress in this area, reconstructing networks with stochastic dynamical processes from limited time series remains to be an outstanding problem. Here we develop a framework based on compressed sensing to reconstruct complex networks on which stochastic spreading dynamics take place. We apply the methodology to a large number of model and real networks, finding that a full reconstruction of inhomogeneous interactions can be achieved from small amounts of polarized (binary) data, a virtue of compressed sensing. Further, we demonstrate that a hidden source that triggers the spreading process but is externally inaccessible can be ascertained and located with high confidence in the absence of direct routes of propagation from it. Our approach thus establishes a paradigm for tracing and controlling epidemic invasion and information diffusion in complex networked systems.

physics.soc-ph↗

arXiv2014

Characterizing and Modeling the Dynamics of Activity and Popularity

Social media, regarded as two-layer networks consisting of users and items, turn out to be the most important channels for access to massive information in the era of Web 2.0. The dynamics of human activity and item popularity is a crucial issue in social media networks. In this paper, by analyzing the growth of user activity and item popularity in four empirical social media networks, i.e., Amazon, Flickr, Delicious and Wikipedia, it is found that cross links between users and items are more likely to be created by active users and to be acquired by popular items, where user activity and item popularity are measured by the number of cross links associated with users and items. This indicates that users generally trace popular items, overall. However, it is found that the inactive users more severely trace popular items than the active users. Inspired by empirical analysis, we propose an evolving model for such networks, in which the evolution is driven only by two-step random walk. Numerical experiments verified that the model can qualitatively reproduce the distributions of user activity and item popularity observed in empirical networks. These results might shed light on the understandings of micro dynamics of activity and popularity in social media networks.

physics.soc-ph↗

arXiv2013

Phase transitions in Ising model induced by weight redistribution on weighted regular networks

In order to investigate the role of the weight in weighted networks, the collective behavior of the Ising system on weighted regular networks is studied by numerical simulation. In our model, the coupling strength between spins is inversely proportional to the corresponding weighted shortest distance. Disordering link weights can effectively affect the process of phase transition even though the underlying binary topological structure remains unchanged. Specifically, based on regular networks with homogeneous weights initially, randomly disordering link weights will change the critical temperature of phase transition. The results suggest that the redistribution of link weights may provide an additional approach to optimize the dynamical behaviors of the system.

cond-mat.stat-mech↗

arXiv2013

Do scientists trace hot topics?

Do scientists follow hot topics in their scientific investigations? In this paper, by performing analysis to papers published in the American Physical Society (APS) Physical Review journals, it is found that papers are more likely to be attracted by hot fields, where the hotness of a field is measured by the number of papers belonging to the field. This indicates that scientists generally do follow hot topics. However, there are qualitative differences among scientists from various countries, among research works regarding different number of authors, different number of affiliations and different number of references. These observations could be valuable for policy makers when deciding research funding and also for individual researchers when searching for scientific projects.

physics.soc-ph↗

arXiv2013

Efficient learning strategy of Chinese characters based on network approach

Based on network analysis of hierarchical structural relations among Chinese characters, we develop an efficient learning strategy of Chinese characters. We regard a more efficient learning method if one learns the same number of useful Chinese characters in less effort or time. We construct a node-weighted network of Chinese characters, where character usage frequencies are used as node weights. Using this hierarchical node-weighted network, we propose a new learning method, the distributed node weight (DNW) strategy, which is based on a new measure of nodes' importance that takes into account both the weight of the nodes and the hierarchical structure of the network. Chinese character learning strategies, particularly their learning order, are analyzed as dynamical processes over the network. We compare the efficiency of three theoretical learning methods and two commonly used methods from mainstream Chinese textbooks, one for Chinese elementary school students and the other for students learning Chinese as a second language. We find that the DNW method significantly outperforms the others, implying that the efficiency of current learning methods of major textbooks can be greatly improved.

physics.soc-ph↗

arXiv2012

Resummation of relativistic corrections to exclusive productions of charmonia in e+ e- collisions

We investigate two exclusive processes, e+ e- -> eta_c + gamma and e+ e- -> J/psi + J/psi, at the center-of-momentum energy sqrt{s}=10.58 GeV within the framework of the nonrelativistic QCD factorization approach. A class of relativistic corrections is resummed to all orders in the heavy-quark velocity v and the corrections are large negative. We further improve the prediction by including available QCD next-to-leading-order corrections and the interference between the QCD and relativistic corrections. The prediction for sigma[e+ e- -> eta_c + gamma] is about 50 fb. In the case of e+ e- -> J/psi + J/psi the standard nonrelativistic QCD prediction for the cross section is negative. As an alternative, the vector-meson-dominance approach is employed to compute the photon-fragmentation contribution of the process, which gives the cross section ~1 fb. This is an indication that the uncalculated QCD higher-order corrections may be significant. Our results can be tested against the forthcoming data from Belle II and super B factories.

hep-ph↗

arXiv2012

Higher-order corrections to exclusive production of charmonia at B factories

As a test of the color-singlet mechanism of the nonrelativistic QCD (NRQCD) factorization approach, we consider the exclusive two-quarkonium productions in electron-positron annihilation e+ e- -> eta_c + gamma and e+ e- -> J/psi + J/psi at B factories. The cross sections are computed to the next-to-leading order in alpha_s and are resummed to all orders in half the relative velocity v of the charm quark in each meson rest frame. The available theoretical prediction of the cross section for e+ e- -> J/psi + eta_c at the same level of theoretical accuracies is consistent with the available experimental data. Those for e+ e- -> eta_c + gamma and e+ e- -> J/psi + J/psi that are computed new in this work can be tested against the data from future super B factories.

hep-ph↗

arXiv2012

Spectral coarse graining for random walk in bipartite networks

Many real-world networks display a natural bipartite structure, while analyzing or visualizing large bipartite networks is one of the most challenges. As a result, it is necessary to reduce the complexity of large bipartite systems and preserve the functionality at the same time. We observe, however, the existing coarse graining methods for binary networks fail to work in the bipartite networks. In this paper, we use the spectral analysis to design a coarse graining scheme specifically for bipartite networks and keep their random walk properties unchanged. Numerical analysis on artificial and real-world bipartite networks indicates that our coarse graining scheme could obtain much smaller networks from large ones, keeping most of the relevant spectral properties. Finally, we further validate the coarse graining method by directly comparing the mean first passage time between the original network and the reduced one.

physics.soc-ph↗

arXiv2012

B-meson Semi-inclusive Decay to $2^{-+}$ Charmonium in NRQCD and X(3872)

The semi-inclusive B-meson decay into spin-singlet D-wave $2^{-+}$ charmonium, $B\to η_{c2}+X$, is studied in nonrelativistic QCD (NRQCD). Both color-singlet and color-octet contributions are calculated at next-to-leading order (NLO) in the strong coupling constant $α_s$. The non-perturbative long-distance matrix elements are evaluated using operator evolution equations. It is found that the color-singlet $^1D_2$ contribution is tiny, while the color-octet channels make dominant contributions. The estimated branching ratio $B(B\to η_{c2}+X)$ is about $0.41\,\times10^{-4}$ in the Naive Dimensional Regularization (NDR) scheme and $1.24\,\times10^{-4}$ in the t'Hooft-Veltman (HV) scheme, with renormalization scale $μ=m_b=4.8$\,GeV. The scheme-sensitivity of these numerical results is due to cancelation between ${}^1S_0^{[8]}$ and ${}^1P_1^{[8]}$ contributions. The $μ$-dependence curves of NLO branching ratios in both schemes are also shown, with $μ$ varying from $\frac{m_b}{2}$ to $2m_b$ and the NRQCD factorization or renormalization scale $μ_Λ$ taken to be $2m_c$. Comparison of the estimated branching ratio of $B\to η_{c2}+X$ with the observed branching ratio of $B \to X(3872)+K$ may lead to the conclusion that X(3872) is unlikely to be the $2^{-+}$ charmonium state $η_{c2}$.

hep-ph↗

arXiv2011

Detecting Important Nodes to Community Structure Using the Spectrum of the Graph

Many complex systems can be represented as networks, and how a network breaks up into subnetworks or communities is of wide interest. However, the development of a method to detect nodes important to communities that is both fast and accurate is a very challenging and open problem. In this manuscript, we introduce a new approach to characterize the node importance to communities. First, a centrality metric is proposed to measure the importance of network nodes to community structure using the spectrum of the adjacency matrix. We define the node importance to communities as the relative change in the eigenvalues of the network adjacency matrix upon their removal. Second, we also propose an index to distinguish two kinds of important nodes in communities, i.e., "community core" and "bridge". Our indices are only relied on the spectrum of the graph matrix. They are applied in many artificial networks as well as many real-world networks. This new methodology gives us a basic approach to solve this challenging problem and provides a realistic result.

physics.soc-ph↗

arXiv2011

Onset of Synchronization in Weighted Complex Networks: the Effect of Weight-Degree Correlation

By numerical simulations, we investigate the onset of synchronization of networked phase oscillators under two different weighting schemes. In scheme-I, the link weights are correlated to the product of the degrees of the connected nodes, so this kind of networks is named as the weight-degree correlated (WDC) network. In scheme-II, the link weights are randomly assigned to each link regardless of the node degrees, so this kind of networks is named as the weight-degree uncorrelated (WDU) network. Interestingly, it is found that by increasing a parameter that governs the weight distribution, the onset of synchronization in WDC network is monotonically enhanced, while in WDU network there is a reverse in the synchronization performance. We investigate this phenomenon from the viewpoint of gradient network, and explain the contrary roles of coupling gradient on network synchronization: gradient promotes synchronization in WDC network, while deteriorates synchronization in WDU network. The findings highlight the fact that, besides the link weight, the correlation between the weight and node degree is also important to the network dynamics.

nlin.CD↗

arXiv2011

Detecting the optimal number of communities in complex networks

To obtain the optimal number of communities is an important problem in detecting community structure. In this paper, we extend the measurement of community detecting algorithms to find the optimal community number. Based on the normalized mutual information index, which has been used as a measure for similarity of communities, a statistic $Ω(c)$ is proposed to detect the optimal number of communities. In general, when $Ω(c)$ reaches its local maximum, especially the first one, the corresponding number of communities \emph{c} is likely to be optimal in community detection. Moreover, the statistic $Ω(c)$ can also measure the significance of community structures in complex networks, which has been paid more attention recently. Numerical and empirical results show that the index $Ω(c)$ is effective in both artificial and real world networks.

physics.soc-ph↗

arXiv2011

Navigation in non-uniform density social networks

Recent empirical investigations suggest a universal scaling law for the spatial structure of social networks. It is found that the probability density distribution of an individual to have a friend at distance $d$ scales as $P(d)\propto d^{-1}$. Since population density is non-uniform in real social networks, a scale invariant friendship network(SIFN) based on the above empirical law is introduced to capture this phenomenon. We prove the time complexity of navigation in 2-dimensional SIFN is at most $O(\log^4 n)$. In the real searching experiment, individuals often resort to extra information besides geography location. Thus, real-world searching process may be seen as a projection of navigation in a $k$-dimensional SIFN($k>2$). Therefore, we also discuss the relationship between high and low dimensional SIFN. Particularly, we prove a 2-dimensional SIFN is the projection of a 3-dimensional SIFN. As a matter of fact, this result can also be generated to any $k$-dimensional SIFN.

physics.soc-ph↗

arXiv2010

Enhancing synchronization by directionality in complex networks

We proposed a method called residual edge-betweenness gradient (REBG) to enhance synchronizability of networks by assignment of link direction while keeping network topology and link weight unchanged. Direction assignment has been shown to improve the synchronizability of undirected networks in general, but we find that in some cases incommunicable components emerge and networks fail to synchronize. We show that the REBG method can effectively avoid the synchronization failure ($R=λ_{2}^{r}/λ_{N}^{r}=0$) which occurs in the residual degree gradient (RDG) method proposed in Phys. Rev. Lett. 103, 228702 (2009). Further experiments show that REBG method enhance synchronizability in networks with community structure as compared with the RDG method.

cond-mat.dis-nn↗