arXiv subjects

Hao Liu

Publications and source records attributed to Hao Liu.

At least 631 records · Page 35Linked to original sources

arXiv2019

CIA-Towards a Unified Marketing Optimization Framework for e-Commerce Sponsored Search

As the largest e-commerce platform, Taobao helps advertisers reach billions of search queries each day via sponsored search, which has also contributed considerable revenue to the platform. An efficient bidding strategy to cater to diverse advertiser demands while balancing platform revenue and consumer experience is significant to a healthy and sustainable marketing ecosystem. In this paper we propose \emph{Customer Intelligent Agent (CIA)}, a bidding optimization framework which implements an impression-level bidding to reflect advertisers' conversion willingness and budget control. In this way, CIA is capable of fulfilling various e-commerce advertiser demands on different levels, such as Gross Merchandise Volume optimization, style comparison etc. Additionally, a replay based simulation system is designed to predict the performance of different take-rate. CIA unifies the benefits of three parties in the marketing ecosystem without changing the Generalized Second Price mechanism. Our extensive offline simulations and large-scale online experiments on \emph{Taobao Search Advertising (TSA)} platform verify the high effectiveness of the CIA framework. Moreover, CIA has been deployed online as a major bidding tool in TSA.

cs.GT↗

arXiv2019

Estimating Individual Advertising Effect in E-Commerce

Online advertising has been the major monetization approach for Internet companies. Advertisers invest budgets to bid for real-time impressions to gain direct and indirect returns. Existing works have been concentrating on optimizing direct returns brought by advertising traffic. However, indirect returns induced by advertising traffic such as influencing the online organic traffic and offline mouth-to-mouth marketing provide extra significant motivation to advertisers. Modeling and quantization of causal effects between the overall advertising return and budget enable the advertisers to spend their money more judiciously. In this paper, we model the overall return as individual advertising effect in causal inference with multiple treatments and bound the expected estimation error with learnable factual loss and distance of treatment-specific context distributions. Accordingly, a representation and hypothesis network is used to minimize the loss bound. We apply the learned causal effect in the online bidding engine of an industry-level sponsored search system. Online experiments show that the causal inference based bidding outperforms the existing online bidding algorithm.

cs.GT↗

arXiv2019

Competitive Experience Replay

Deep learning has achieved remarkable successes in solving challenging reinforcement learning (RL) problems when dense reward function is provided. However, in sparse reward environment it still often suffers from the need to carefully shape reward function to guide policy optimization. This limits the applicability of RL in the real world since both reinforcement learning and domain-specific knowledge are required. It is therefore of great practical importance to develop algorithms which can learn from a binary signal indicating successful task completion or other unshaped, sparse reward signals. We propose a novel method called competitive experience replay, which efficiently supplements a sparse reward by placing learning in the context of an exploration competition between a pair of agents. Our method complements the recently proposed hindsight experience replay (HER) by inducing an automatic exploratory curriculum. We evaluate our approach on the tasks of reaching various goal locations in an ant maze and manipulating objects with a robotic arm. Each task provides only binary rewards indicating whether or not the goal is achieved. Our method asymmetrically augments these sparse rewards for a pair of agents each learning the same task, creating a competitive game designed to drive exploration. Extensive experiments demonstrate that this method leads to faster converge and improved task performance.

cs.LG↗

arXiv2019

Large-scale Bisample Learning on ID Versus Spot Face Recognition

In real-world face recognition applications, there is a tremendous amount of data with two images for each person. One is an ID photo for face enrollment, and the other is a probe photo captured on spot. Most existing methods are designed for training data with limited breadth (a relatively small number of classes) and sufficient depth (many samples for each class). They would meet great challenges on ID versus Spot (IvS) data, including the under-represented intra-class variations and an excessive demand on computing devices. In this paper, we propose a deep learning based large-scale bisample learning (LBL) method for IvS face recognition. To tackle the bisample problem with only two samples for each class, a classification-verification-classification (CVC) training strategy is proposed to progressively enhance the IvS performance. Besides, a dominant prototype softmax (DP-softmax) is incorporated to make the deep learning scalable on large-scale classes. We conduct LBL on a IvS face dataset with more than two million identities. Experimental results show the proposed method achieves superior performance to previous ones, validating the effectiveness of LBL on IvS face recognition.

cs.CV↗

arXiv2018

Sequence-based Person Attribute Recognition with Joint CTC-Attention Model

Attribute recognition has become crucial because of its wide applications in many computer vision tasks, such as person re-identification. Like many object recognition problems, variations in viewpoints, illumination, and recognition at far distance, all make this task challenging. In this work, we propose a joint CTC-Attention model (JCM), which maps attribute labels into sequences to learn the semantic relationship among attributes. Besides, this network uses neural network to encode images into sequences, and employs connectionist temporal classification (CTC) loss to train the network with the aim of improving the encoding performance of the network. At the same time, it adopts the attention model to decode the sequences, which can realize aligning the sequences and better learning the semantic information from attributes. Extensive experiments on three public datasets, i.e., Market-1501 attribute dataset, Duke attribute dataset and PETA dataset, demonstrate the effectiveness of the proposed method.

cs.CV↗

arXiv2018

E and B families of the Stokes parameters in the polarized synchrotron and thermal dust foregrounds

Better understanding of Galactic foregrounds is one of the main obstacles to detection of primordial gravitational waves through measurement of the B mode in the polarized microwave sky. We generalize the method proposed in \cite{2011PhRvD..83h3003B} and decompose the polarization signals into the E and B families directly in the domain of the Stokes $Q$, $U$ parameters as $(Q,U)\equiv(Q_E, U_E)+(Q_B,U_B)$. This also enables an investigation of the morphology and the frequency dependence of these two families, which has been done in the WMAP K, Ka (tracing synchrotron emission) and Planck 2015 HFI maps (tracing thermal dust). The results reveal significant differences in spectra between the E and B families. The spectral index of the E family fluctuates less across the sky than that of the B family, and the same tendency occurs for the polarization angles of the dust and synchrotron channels. The new insight from WMAP and Planck data on the North Polar Spur and BICEP2 zones through our method clearly indicates that these zones are characterized by very low polarization intensity of the B family compared to the E family. We have detected global structure of the B family polarization angles at high Galactic latitudes which cannot be attributed to the cosmic microwave background or instrumental noise. However, we cannot exclude instrumental systematics as a partial contributor to these anomalies.

astro-ph.CO↗

arXiv2018

Closed-form solution for the edge vortex of a revolving plate

Flapping and revolving wings can produce attached leading edge vortices (LEVs) when the angle of attack is large. In this work, a low order model is proposed for the edge vortices that develop on a revolving plate at 90 degrees angle of attack which is the simplest limiting case, yet showing remarkable similarity with the generally known LEVs. The problem is solved analytically, providing short closed-form expressions for the circulation and the position of the vortex. A good agreement with the numerical solution of the Navier-Stokes equations suggests that, for the conditions examined, the vorticity production at the sharp edge and its subsequent three-dimensional transport are the main effects that shape the edge vortex.

physics.flu-dyn↗

arXiv2018

Fingerprint of Galactic Loop I on polarized microwave foregrounds

Context: Currently, detection of the primordial gravitational waves by the B-mode of Cosmic Microwave Background (CMB) is primarily limited by our knowledge of the polarized microwave foreground emissions. Thus improvements of the foreground analysis are necessary. As revealed in~\cite{2018arXiv180410382L}, the E-mode and B-mode of the polarized foreground have noticeable different properties, both in morphology and frequency spectrum, suggesting that they arise from different physical processes, and need to be studied separately. Aims: I will study the polarized emission from Galactic loops, especially Loop I, and mainly focus on the following issues: Does it contribute predominantly to the E-mode or B-mode? In which frequency bands and in which sky regions can it be identified? Methods: Based on a well known result about the magnetic field alignment in supernova explosions, a theoretical expectation is established that the loop polarizations should be predominantly E-mode. In particular, the expected polarization angles of Loop I are compared with those from the real microwave band data of WMAP and Planck. Results and conclusions: The comparison between model and data shows remarkable consistency between data and expectation at all bands and for a large area of the sky. This result suggests that the polarized emission of Galactic Loop I is a major polarized component in all microwave bands from 23 to 353 GHz, and a considerable part of the polarized foreground is likely originated from a local bubble associated with Loop I, instead of the far more distant Galactic emission. The result also provides a possible way to explain the reported E-to-B excess~\citep{2016A&A...586A.133P} by contribution of the loops. Finally, this work may also provide the first geometrical evidence that the Earth was hit by a supernova explosion.

astro-ph.CO↗

arXiv2018

Adaptive guaranteed-performance consensus design for high-order multiagent systems

The current paper addresses the distributed guaranteed-performance consensus design problems for general high-order linear multiagent systems with leaderless and leader-follower structures, respectively. The information about the Laplacian matrix of the interaction topology or its minimum nonzero eigenvalue is usually required in existing works on the guaranteed-performance consensus, which means that their conclusions are not completely distributed. A new translation-adaptive strategy is proposed to realize the completely distributed guaranteed-performance consensus control by using the structure feature of a complete graph in the current paper. For the leaderless case, an adaptive guaranteed-performance consensualization criterion is given in terms of Riccati inequalities and a regulation approach of the consensus control gain is presented by linear matrix inequalities. Extensions to the leader-follower cases are further investigated. Especially, the guaranteed-performance costs for leaderless and leader-follower cases are determined, respectively, which are associated with the intrinsic structure characteristic of the interaction topologies. Finally, two numerical examples are provided to demonstrate theoretical results.

cs.MA↗

arXiv2018

Completely Distributed Guaranteed-performance Consensualization for High-order Multiagent Systems with Switching Topologies

The guaranteed-performance consensualization for high-order linear and nonlinear multiagent systems with switching topologies is respectively realized in a completely distributed manner in the sense that consensus design criteria are independent of interaction topologies and switching motions. The current paper firstly proposes an adaptive consensus protocol with guaranteed-performance constraints and switching topologies, where interaction weights among neighboring agents are adaptively adjusted and state errors among all agents can be regulated. Then, a new translation-adaptive strategy is shown to realize completely distributed guaranteed-performance consensus control and an adaptive guaranteed-performance consensualization criterion is given on the basis of the Riccati inequality. Furthermore, an approach to regulate the consensus control gain and the guaranteed-performance cost is proposed in terms of linear matrix inequalities. Moreover, main conclusions for linear multiagent systems are extended to Lipschitz nonlinear cases. Finally, two numerical examples are provided to demonstrate theoretical results.

eess.SY↗

arXiv2018

Spatial social networks identified from urban group travel

While the individual travel implicates the trace of individual mobility decision, group travels signify the possible social relationship behind these traces. Different from online social network, spatial interaction between individuals is a critical yet unknown dimension to understand the collective behaviors in a city. In this paper, based on over 127 million trips in Beijing metro network, we develop a method to distinguish the group travel of friends from the encounter travel of familiar strangers. We find travels of friends are among the most predictable groups. These identified friendships are interwoven and form a friendship network, with structural properties different from encounter network. The topological role of individuals in this network is found strongly correlated with her travel predictability. The overall time savings of about 34190 minutes after redistribution of inefficient group traveler with flexible travel purposes shows the potential of designing specific traffic fares for group travel. Our identification and understanding of group travel may help to develop and organize new traffic mode in the future smart transportation.

physics.soc-ph↗

arXiv2018

THz pulses over 50 millijoules generated from relativistic picosecond laser-plasma interactions

Ultrahigh-power terahertz (THz) radiation sources are essential for many applications, such as nonlinear THz physics, THz-wave based compact accelerators, etc. However, until now none of THz sources reported, whether based upon large-scale accelerators or high power lasers, have produced THz pulses with energies above the millijoule (mJ) barrier. Here we report on the efficient generation of low-frequency (<3 THz) THz pulses with unprecedentedly high energies over 50 mJ. The THz radiation is produced by coherent transition radiation of a picosecond laser-accelerated ultra-bright bunch of relativistic electrons from a solid target. Such high energy THz pulses can not only trigger various nonlinear dynamics in matter, but also open up a new research field of relativistic THz optics.

physics.plasm-ph↗

arXiv2018

Photometric and Spectroscopic Observations of GRB 140629A

We present our optical photometric and spectroscopical observations of GRB 140629A. A redshift of $z=2.275\pm0.043$ is measured through the metal absorption lines in our spectroscopic data. Using our photometric data and multiple observational data observed with other telescopes, we show that its optical light curve is well interpreted with the standard forward shock models in the thin shell case. Its optical-X-ray afterglow spectrum is jointly fitted with a single power-law function, yielding a photon index of $-1.90\pm 0.05$. The optical extinction and neutral hydrogen absorption of the GRB host galaxy are negligible. The fit to the light curve with the standard models shows that the ambient density is $60\pm 9$ cm$^{-3}$ and the GRB radiating efficiency is as low as $\sim 0.24\%$, likely indicating a baryonic-dominated ejecta of this GRB. This burst agrees well with the $L_{\rm p, iso}-E_p^{'}-Γ_0$ relation, but confidently violates those empirical relations involving geometric corrections (or jet break time). This gives rise to an issue of possible selection effect on these relations since the jet opening angle of this GRB is extremely narrow (0.04 rad).

astro-ph.HE↗

arXiv2018

Degeneracy of gravitational waveforms in the context of GW150914

We study the degeneracy of theoretical gravitational waveforms for binary black hole mergers using an aligned-spin effective-one-body model. After appropriate truncation, bandpassing, and matching, we identify regions in the mass--spin parameter space containing waveforms similar to the template proposed for GW150914, with masses $m_1 = 36^{+5}_{-4} M_\odot$ and $m_2 = 29^{+4}_{-4} M_\odot$, using the cross-correlation coefficient as a measure of the similarity between waveforms. Remarkably high cross-correlations are found across broad regions of parameter space. The associated uncertanties exceed these from LIGO's Bayesian analysis considerably. We have shown that waveforms with greatly increased masses, such as $m_1 = 70 M_\odot$ and $m_2 = 35 M_\odot$, and strong anti-aligned spins ($χ_1=0.95$ and $χ_2=-0.95$) yield almost the same signal-to-noise ratio in the strain data for GW150914.

gr-qc↗

arXiv2018

Multi-View Image Generation from a Single-View

This paper addresses a challenging problem -- how to generate multi-view cloth images from only a single view input. To generate realistic-looking images with different views from the input, we propose a new image generation model termed VariGANs that combines the strengths of the variational inference and the Generative Adversarial Networks (GANs). Our proposed VariGANs model generates the target image in a coarse-to-fine manner instead of a single pass which suffers from severe artifacts. It first performs variational inference to model global appearance of the object (e.g., shape and color) and produce a coarse image with a different view. Conditioned on the generated low resolution images, it then proceeds to perform adversarial learning to fill details and generate images of consistent details with the input. Extensive experiments conducted on two clothing datasets, MVC and DeepFashion, have demonstrated that images of a novel view generated by our model are more plausible than those generated by existing approaches, in terms of more consistent global appearance as well as richer and sharper details.

cs.CV↗

arXiv2018

Action-depedent Control Variates for Policy Optimization via Stein's Identity

Policy gradient methods have achieved remarkable successes in solving challenging reinforcement learning problems. However, it still often suffers from the large variance issue on policy gradient estimation, which leads to poor sample efficiency during training. In this work, we propose a control variate method to effectively reduce variance for policy gradient methods. Motivated by the Stein's identity, our method extends the previous control variate methods used in REINFORCE and advantage actor-critic by introducing more general action-dependent baseline functions. Empirical studies show that our method significantly improves the sample efficiency of the state-of-the-art policy gradient approaches.

stat.ML↗

arXiv2018

Dynamic Output Feedback Guaranteed-Cost Synchronization for Multiagent Networks with Given Cost Budgets

The current paper addresses the distributed guaranteed-cost synchronization problems for general high-order linear multiagent networks. Existing works on the guaranteed-cost synchronization usually require all state information of neighboring agents and cannot give the cost budget previously. For both leaderless and leader-following interaction topologies, the current paper firstly proposes a dynamic output feedback synchronization protocol with guaranteed-cost constraints, which can realize the tradeoff design between the energy consumption and the synchronization regulation performance with the given cost budget. Then, according to different structure features of interaction topologies, leaderless and leader-following guaranteed-cost synchronization analysis and design criteria are presented, respectively, and an algorithm is proposed to deal with the impacts of nonlinear terms by using both synchronization analysis and design criteria. Especially, an explicit expression of the synchronization function is shown for leaderless cases, which is independent of protocol states and the given cost budget. Finally, numerical examples are presented to demonstrate theoretical results.

eess.SY↗

arXiv2018

Guaranteed-cost consensus for multiagent networks with Lipschitz nonlinear dynamics and switching topologies

Guaranteed-cost consensus for high-order nonlinear multi-agent networks with switching topologies is investigated. By constructing a time-varying nonsingular matrix with a specific structure, the whole dynamics of multi-agent networks is decomposed into the consensus and disagreement parts with nonlinear terms, which is the key challenge to be dealt with. An explicit expression of the consensus dynamics, which contains the nonlinear term, is given and its initial state is determined. Furthermore, by the structure property of the time-varying nonsingular transformation matrix and the Lipschitz condition, the impacts of the nonlinear term on the disagreement dynamics are linearized and the gain matrix of the consensus protocol is determined on the basis of the Riccati equation. Moreover, an approach to minimize the guaranteed cost is given in terms of linear matrix inequalities. Finally, the numerical simulation is shown to demonstrate the effectiveness of theoretical results.

eess.SY↗