• ### Matrix Product Operators for Sequence to Sequence Learning(1803.10908)

The method of choice to study one-dimensional strongly interacting many body quantum systems is based on matrix product states and operators. Such method allows to explore the most relevant, and numerically manageable, portion of an exponentially large space. It also allows to describe accurately correlations between distant parts of a system, an important ingredient to account for the context in machine learning tasks. Here we introduce a machine learning model in which matrix product operators are trained to implement sequence to sequence prediction, i.e. given a sequence at a time step, it allows one to predict the next sequence. We then apply our algorithm to cellular automata (for which we show exact analytical solutions in terms of matrix product operators), and to nonlinear coupled maps. We show advantages of the proposed algorithm when compared to conditional random fields and bidirectional long short-term memory neural network. To highlight the flexibility of the algorithm, we also show that it can readily perform classification tasks.
• ### Improving the Improved Training of Wasserstein GANs: A Consistency Term and Its Dual Effect(1803.01541)

March 5, 2018 cs.CV, cs.LG, stat.ML
Despite being impactful on a variety of problems and applications, the generative adversarial nets (GANs) are remarkably difficult to train. This issue is formally analyzed by \cite{arjovsky2017towards}, who also propose an alternative direction to avoid the caveats in the minmax two-player training of GANs. The corresponding algorithm, called Wasserstein GAN (WGAN), hinges on the 1-Lipschitz continuity of the discriminator. In this paper, we propose a novel approach to enforcing the Lipschitz continuity in the training procedure of WGANs. Our approach seamlessly connects WGAN with one of the recent semi-supervised learning methods. As a result, it gives rise to not only better photo-realistic samples than the previous methods but also state-of-the-art semi-supervised learning results. In particular, our approach gives rise to the inception score of more than 5.0 with only 1,000 CIFAR-10 images and is the first that exceeds the accuracy of 90% on the CIFAR-10 dataset using only 4,000 labeled images, to the best of our knowledge.
• ### Relativistic, single-cycle tunable-infrared pulses generated by a tailored plasma density structure(1712.09182)

The availability of intense, ultrashort coherent radiation sources in the infrared region of the spectrum is enabling the generation of attosecond X-ray pulses via high harmonic generation, pump-probe experiments in the "molecular fingerprint" region and opening up the area of relativistic-infrared nonlinear optics of plasmas. These applications would benefit from multi-millijoule single-cycle pulses in the mid to long wavelength infrared (LW-IR) region. Here we present a new scheme capable of producing tunable relativistically intense, single-cycle infrared pulses from 5-14$\mu$m with a 1.7% conversion efficiency based on a photon frequency downshifting scheme that uses a tailored plasma density structure. The carrier-envelope phase (CEP) of the LW-IR pulse is locked to that of the drive laser to within a few percent. Such a versatile tunable IR source may meet the demands of many cutting-edge applications in strong-field physics and greatly promote their development.
• ### A Clifford Algebra Approach to Chiral Symmetry Breaking and Fermion Mass Hierarchies(1709.09294)

Sept. 27, 2017 hep-th, hep-ph
We propose a Clifford algebra approach to chiral symmetry breaking and fermion mass hierarchies in the context of composite Higgs bosons. Standard model fermions are represented by algebraic spinors of six-dimensional binary Clifford algebra, while ternary Clifford algebra-related flavor projection operators control allowable flavor-mixing interactions. There are three composite electroweak Higgs bosons resulted from top quark, tau neutrino, and tau lepton condensations. Each of the three condensations gives rise to masses of four different fermions. The fermion mass hierarchies within these three groups are determined by four-fermion condensations, which break two global chiral symmetries. The four-fermion condensations induce axion-like pseudo-Nambu-Goldstone bosons and can be dark matter candidates. In addition to the 125 GeV Higgs boson observed at the Large Hadron Collider, we anticipate detection of tau neutrino composite Higgs boson via the charm quark decay channel.
• ### Dynamical Anisotropic Response of Black Phosphorus under Magnetic Field(1709.03079)

Sept. 10, 2017 cond-mat.mtrl-sci
Black phosphorus (BP) has emerged as a promising material candidate for next generation electronic and optoelectronic devices due to its high mobility, tunable band gap and highly anisotropic properties. In this work, polarization resolved ultrafast mid-infrared transient reflection spectroscopy measurements are performed to study the dynamical anisotropic optical properties of BP under magnetic fields up to 9 T. The relaxation dynamics of photoexcited carrier is found to be insensitive to the applied magnetic field due to the broadening of the Landau levels and large effective mass of carriers. While the anisotropic optical response of BP decreases with increasing magnetic field, its enhancement due to the excitation of hot carriers is similar to that without magnetic field. These experimental results can be well interpreted by the magneto-optical conductivity of the Landau levels of BP thin film, based on an effective k*p Hamiltonian and linear response theory. These findings suggest attractive possibilities of multi-dimensional controls of anisotropic response (AR) of BP with light, electric and magnetic field, which further introduces BP to the fantastic magnetic field sensitive applications.
• ### Digital image splicing detection based on Markov features in QDCT and QWT domain(1708.08245)

Aug. 28, 2017 cs.CR, cs.CV
Image splicing detection is of fundamental importance in digital forensics and therefore has attracted increasing attention recently. In this paper, a color image splicing detection approach is proposed based on Markov transition probability of quaternion component separation in quaternion discrete cosine transform (QDCT) domain and quaternion wavelet transform (QWT) domain. Firstly, Markov features of the intra-block and inter-block between block QDCT coefficients are obtained from the real part and three imaginary parts of QDCT coefficients respectively. Then, additional Markov features are extracted from luminance (Y) channel in quaternion wavelet transform domain to characterize the dependency of position among quaternion wavelet subband coefficients. Finally, ensemble classifier (EC) is exploited to classify the spliced and authentic color images. The experiment results demonstrate that the proposed approach can outperforms some state-of-the-art methods.
• ### Refutations on "Debunking the Myths of Influence Maximization: An In-Depth Benchmarking Study"(1705.05144)

June 9, 2017 cs.SI
In a recent SIGMOD paper titled "Debunking the Myths of Influence Maximization: An In-Depth Benchmarking Study", Arora et al. [1] undertake a performance benchmarking study of several well-known algorithms for influence maximization. In the process, they contradict several published results, and claim to have unearthed and debunked several "myths" that existed around the research of influence maximization. It is the goal of this article to examine their claims objectively and critically, and refute the erroneous ones. Our investigation discovers that first, the overall experimental methodology in Arora et al. [1] is flawed and leads to scientifically incorrect conclusions. Second, the paper [1] is riddled with issues specific to a variety of influence maximization algorithms, including buggy experiments, and draws many misleading conclusions regarding those algorithms. Importantly, they fail to appreciate the trade-off between running time and solution quality, and did not incorporate it correctly in their experimental methodology. In this article, we systematically point out the issues present in [1] and refute 11 of their misclaims.
• ### Room-temperature high detectivity mid-infrared photodetectors based on black arsenic phosphorus(1705.00801)

The mid-infrared (MIR) spectral range, pertaining to important applications such as molecular 'fingerprint' imaging, remote sensing, free space telecommunication and optical radar, is of particular scientific interest and technological importance. However, state-of-the-art materials for MIR detection are limited by intrinsic noise and inconvenient fabrication processes, resulting in high cost photodetectors requiring cryogenic operation. We report black arsenic-phosphorus-based long wavelength infrared photodetectors with room temperature operation up to 8.2 um, entering the second MIR atmospheric transmission window. Combined with a van der Waals heterojunction, room temperature specific detectivity higher than 4.9*10^9 Jones was obtained in the 3-5 um range. The photodetector works in a zero-bias photovoltaic mode, enabling fast photoresponse and low dark noise. Our van der Waals heterojunction photodector not only exemplify black arsenic-phosphorus as a promising candidate for MIR opto-electronic applications, but also pave the way for a general strategy to suppress 1/f noise in photonic devices.
• ### Revenue Maximization in Incentivized Social Advertising(1612.00531)

April 19, 2017 cs.SI, cs.DB
• ### Fixed versus Dynamic Co-Occurrence Windows in TextRank Term Weights for Information Retrieval(1704.01851)

April 6, 2017 cs.IR
TextRank is a variant of PageRank typically used in graphs that represent documents, and where vertices denote terms and edges denote relations between terms. Quite often the relation between terms is simple term co-occurrence within a fixed window of k terms. The output of TextRank when applied iteratively is a score for each vertex, i.e. a term weight, that can be used for information retrieval (IR) just like conventional term frequency based term weights. So far, when computing TextRank term weights over co- occurrence graphs, the window of term co-occurrence is al- ways ?xed. This work departs from this, and considers dy- namically adjusted windows of term co-occurrence that fol- low the document structure on a sentence- and paragraph- level. The resulting TextRank term weights are used in a ranking function that re-ranks 1000 initially returned search results in order to improve the precision of the ranking. Ex- periments with two IR collections show that adjusting the vicinity of term co-occurrence when computing TextRank term weights can lead to gains in early precision.
• ### Rhetorical relations for information retrieval(1704.01599)

April 5, 2017 cs.CL, cs.IR
Typically, every part in most coherent text has some plausible reason for its presence, some function that it performs to the overall semantics of the text. Rhetorical relations, e.g. contrast, cause, explanation, describe how the parts of a text are linked to each other. Knowledge about this socalled discourse structure has been applied successfully to several natural language processing tasks. This work studies the use of rhetorical relations for Information Retrieval (IR): Is there a correlation between certain rhetorical relations and retrieval performance? Can knowledge about a document's rhetorical relations be useful to IR? We present a language model modification that considers rhetorical relations when estimating the relevance of a document to a query. Empirical evaluation of different versions of our model on TREC settings shows that certain rhetorical relations can benefit retrieval effectiveness notably (> 10% in mean average precision over a state-of-the-art baseline).
• ### Ultrafast Relaxation Dynamics of Photoexcited Dirac Fermion in The Three Dimensional Dirac Semimetal Cadmium Arsenide(1608.07361)

Three dimensional (3D) Dirac semimetals which can be seen as 3D analogues of graphene have attracted enormous interests in research recently. In order to apply these ultrahigh-mobility materials in future electronic/optoelectronic devices, it is crucial to understand the relaxation dynamics of photoexcited carriers and their coupling with lattice. In this work, we report ultrafast transient reflection measurements of the photoexcited carrier dynamics in cadmium arsenide (Cd3As2), which is one of the most stable Dirac semimetals that have been confirmed experimentally. By using low energy probe photon of 0.3 eV, we probed the dynamics of the photoexcited carriers that are Dirac-Fermi-like approaching the Dirac point. We systematically studied the transient reflection on bulk and nanoplate samples that have different doping intensities by tuning the probe wavelength, pump power and lattice temperature, and find that the dynamical evolution of carrier distributions can be retrieved qualitatively by using a two-temperature model. This result is very similar to that of graphene, but the carrier cooling through the optical phonon couplings is slower and lasts over larger electron temperature range because the optical phonon energies in Cd3As2 are much lower than those in graphene.
• ### Imaging non-local electron transport via local excess noise(1610.01711)

Oct. 23, 2016 cond-mat.mes-hall
Noise is usually a hindrance to signal detection. As stressed by Landauer, however, noise can be an invaluable signal that reveals kinetics of charge particles. Understanding local non-equilibrium electron kinetics at nano-scale is of decisive importance for the development of miniaturized electronic devices, optical nano-devices, and heat management devices. In non-equilibrium conditions electrons cause current fluctuation (excess noise) that contains fingerprint-like information about the electron kinetics. A crucial challenge is hence a local detection of excess noise and its real-space mapping. However, the challenge has not been tackled in existing noise measurements because the noise studied was the spatially integrated one. Here we report the experiment in which the excess noise at ultra-high-frequency(21.3THz), generated on GaAs/AlGaAs quantum well (QW) devices with a nano-scale constriction, is locally detected and mapped for the first time. We use a sharp tungsten tip as a movable, contact-free and noninvasive probe of the local noise, and achieved nano-scale spatial resolution (~50nm). Local profile of electron heating and hot-electron kinetics at nano-scales are thereby visualized for the first time, disclosing remarkable non-local nature of the transport, stemming from the velocity overshoot and the intervalley hot electron transfer. While we demonstrate the usefulness of our experimental method by applying to mesoscopic conductors, we emphasize that the method is applicable to a variety of different materials beyond the conductor, and term our instrument a scanning noise microscope (SNoiM):In general non-equilibrium current fluctuations are generated in any materials including dielectrics, metals and molecular systems. The fluctuations, in turn, excite fluctuating electric and magnetic evanescent fields on the material surface, which can be detected and imaged by our SNoiM.
• ### A Study of Factuality, Objectivity and Relevance: Three Desiderata in Large-Scale Information Retrieval?(1610.01327)

Oct. 8, 2016 cs.IR
Much of the information processed by Information Retrieval (IR) systems is unreliable, biased, and generally untrustworthy [1], [2], [3]. Yet, factuality & objectivity detection is not a standard component of IR systems, even though it has been possible in Natural Language Processing (NLP) in the last decade. Motivated by this, we ask if and how factuality & objectivity detection may benefit IR. We answer this in two parts. First, we use state-of-the-art NLP to compute the probability of document factuality & objectivity in two TREC collections, and analyse its relation to document relevance. We find that factuality is strongly and positively correlated to document relevance, but objectivity is not. Second, we study the impact of factuality & objectivity to retrieval effectiveness by treating them as query independent features that we combine with a competitive language modelling baseline. Experiments with 450 TREC queries show that factuality improves precision >10% over strong baselines, especially for uncurated data used in web search; objectivity gives mixed results. An overall clear trend is that document factuality & objectivity is much more beneficial to IR when searching uncurated (e.g. web) documents vs. curated (e.g. state documentation and newswire articles). To our knowledge, this is the first study of factuality & objectivity for back-end IR, contributing novel findings about the relation between relevance and factuality/objectivity, and statistically significant gains to retrieval effectiveness in the competitive web search task.
• ### Ultrafast Broadband Photodetectors based on Three-dimensional Dirac Semimetal Cd3As2(1609.00768)

Sept. 2, 2016 cond-mat.mes-hall
The efforts to pursue photo detection with extreme performance in terms of ultrafast response time, broad detection wavelength range, and high sensitivity have never been exhausted as driven by its wide range of optoelectronic and photonic applications such as optical communications, interconnects, imaging and remote sensing1. 2D Dirac semimetal graphene has shown excellent potential toward high performance photodetector with high operation speed, broadband response and efficient carrier multiplications benefiting from its linear dispersion band structure with high carrier mobility and zero bandgap2-4. As the three dimensional analogues of graphene, Dirac semimetal Cd3As2 processes all advantages of graphene as a photosensitive material but potentially has stronger interaction with light as bulk material and thus enhanced responsivity5,6, which promises great potential in improving the performance of photodetector in various aspects . In this work, we report the realization of an ultrafast broadband photodetector based on Cd3As2. The prototype metal-Cd3As2-metal photodetector exhibits a responsivity of 5.9 mA/W with response time of about 6.9 ps without any special device optimization. Broadband responses from 0.8 eV to 2.34 eV are measured with potential detection range extendable to far infrared and terahertz. Systematical studies indicate that the photo-thermoelectric effect plays important roles in photocurrent generation, similar to that in graphene. Our results suggest this emerging class of exotic quantum materials can be harnessed for photo detection with high sensitivity and high speed (~145 GHz) in challenging middle/far-infrared and THz range.
• ### Broadband Circular Polarizers Constructed by Helix-like Chiral Metamaterials(1606.05399)

June 17, 2016 physics.optics
In this paper, a kind of helix-like chiral metamaterial, which can be realized with multiple conventional lithography or electron beam lithographic techniques, is proposed to achieve broadband bianisotropic optical response analogous to helical metamaterial. On the basis of twisted metamaterial, via tailoring the relative orientation within the lattice, the anisotropy of arc is converted into magneto-electric coupling of closely spaced arc pairs, which leads to a broad bianisotropic optical response. By connecting the adjacent upper and lower arcs, the coupling of metasurface pairs is transformed to the coupling of the three-dimensional inclusions, and provides a much broader and higher bianisotropic optical response. For only a four-layer helix-like metamaterial, the maximum extinction ratio can reach 19.7. The operation band is in the wavelength range from 4.69 {\mu}m to 8.98 {\mu}m with an average extinction ratio of 6.9. And the transmittance for selective polarization is above 0.8 in the entire operation band. Such a structure is promising for integratable and scalable broadband circular polarizers, especially has great potential to act as broadband circular micropolarizers in the field of the full-stokes division of focal plane polarimeters.
• ### Interference-aided spectrum fitting method for accurately film thickness determination(1606.05402)

June 17, 2016 physics.optics
A new approach was proposed to accurately determine the thickness of film, especially for ultra-thin film, through spectrum fitting with the assistance of interference layer. The determination limit can reach even less than 1 nm. Its accuracy is far better than traditional methods. This determination method is verified by experiments and the determination limit is at least 3.5 nm compared with the results of AFM. Furthermore, double-interference-aided spectra fitting method is proposed to reduce the requirements of determination instruments, which allow one to determine the film thickness with a low precision common spectrometer and largely lower the cost. It is a very high precision determination method for on-site and in-situ applications, especially for ultra-thin films.
• ### Electroweak and Majorana Sector Higgs Bosons and Pseudo-Nambu-Goldstone Bosons(1603.04697)

June 6, 2016 hep-ph, physics.gen-ph
We propose a Clifford algebra based model, which treats both gravity and Yang-Mills interactions as gauge fields. There are two sectors of boson fields as electroweak and Majorana bosons. The electroweak boson sector induces fermion masses via spontaneous symmetry breaking. It is composed of scalar Higgs, pseudoscalar Higgs, and antisymmetric tensor components. The Majorana boson sector contributes to flavor mixing and Majorana masses of right-handed neutrinos. It is comprised of neutrino Higgs and pseudo-Nambu-Goldstone bosons. The LHC 750 GeV diphoton resonance might possibly be identified as a Majorana sector pseudo-Nambu-Goldstone boson, which results from spontaneous symmetry breaking of a flavor-related global U(1) symmetry involving four-fermion condensation of right-handed leptons and quarks. The diphoton decay is loop induced, since tree-level decay is suppressed by large Majorana mass of the right-handed neutrino. There is also a potential dark matter candidate, which is the four-lepton condensation of muon, muon-neutrino, tau, and tau-neutrino.
• ### Controlling the Numerical Cerenkov Instability in PIC simulations using a customized finite difference Maxwell solver and a local FFT based current correction(1605.01496)

May 5, 2016 physics.comp-ph
In this paper we present a customized finite-difference-time-domain (FDTD) Maxwell solver for the particle-in-cell (PIC) algorithm. The solver is customized to effectively eliminate the numerical Cerenkov instability (NCI) which arises when a plasma (neutral or non-neutral) relativistically drifts on a grid when using the PIC algorithm. We control the EM dispersion curve in the direction of the plasma drift of a FDTD Maxwell solver by using a customized higher order finite difference operator for the spatial derivative along the direction of the drift ($\hat 1$ direction). We show that this eliminates the main NCI modes with moderate $\vert k_1 \vert$, while keeps additional main NCI modes well outside the range of physical interest with higher $\vert k_1 \vert$. These main NCI modes can be easily filtered out along with first spatial aliasing NCI modes which are also at the edge of the fundamental Brillouin zone. The customized solver has the possible advantage of improved parallel scalability because it can be easily partitioned along $\hat 1$ which typically has many more cells than other directions for the problems of interest. We show that FFTs can be performed locally to current on each partition to filter out the main and first spatial aliasing NCI modes, and to correct the current so that it satisfies the continuity equation for the customized spatial derivative. This ensures that Gauss' Law is satisfied. We present simulation examples of one relativistically drifting plasmas, of two colliding relativistically drifting plasmas, and of nonlinear laser wakefield acceleration (LWFA) in a Lorentz boosted frame that show no evidence of the NCI can be observed when using this customized Maxwell solver together with its NCI elimination scheme.
• ### Band gap modulation in {\gamma}-graphyne by p-n codoping(1604.04370)

The modulation of band gap in the two-dimensional carbon materials is of impor- tance for their applications as electronic devices. By first-principles calculations, we propose a model to control the band gap size of {\gamma}-graphyne. The model is named as p-n codoping, i. e., using B and N atoms to codope into {\gamma}-graphyne. After codoping, B atom plays a role of p doping and N atom acts as n doping. The Fermi energy level returns around the forbidden zone and the band gap of {\gamma}-graphyne vary bigger or smaller. Moreover, the gaps exhibit an oscillated behaviour in different codoping configurations. The proposed model serves as new insights for better modulation of the electronic properties of 2D carbon materials.
• ### The Intensity Matching Approach: A Tractable Stochastic Geometry Approximation to System-Level Analysis of Cellular Networks(1604.02683)

April 10, 2016 cs.IT, math.IT
The intensity matching approach for tractable performance evaluation and optimization of cellular networks is introduced. It assumes that the base stations are modeled as points of a Poisson point process and leverages stochastic geometry for system-level analysis. Its rationale relies on observing that system-level performance is determined by the intensity measure of transformations of the underlaying spatial Poisson point process. By approximating the original system model with a simplified one, whose performance is determined by a mathematically convenient intensity measure, tractable yet accurate integral expressions for computing area spectral efficiency and potential throughput are provided. The considered system model accounts for many practical aspects that, for tractability, are typically neglected, e.g., line-of-sight and non-line-of-sight propagation, antenna radiation patterns, traffic load, practical cell associations, general fading channels. The proposed approach, more importantly, is conveniently formulated for unveiling the impact of several system parameters, e.g., the density of base stations and blockages. The effectiveness of this novel and general methodology is validated with the aid of empirical data for the locations of base stations and for the footprints of buildings in dense urban environments.
• ### The mechanism and process of spontaneous boron doping in graphene in the theoretical perspective(1602.08677)

Feb. 28, 2016 cond-mat.mes-hall
A theoretical model is presented to reveal the mechanism of B doping into graphene in the microwave plasma experiment choosing trimethylboron as the doping source (ACS NANO 6 (2012) 1970). The results show that the reason for B doping comes from the combinational interaction of B and other groups (C, H, CH, CH2 or CH3) decomposing from trimethylboron and the doping undergoes two crucial steps. The minimal energy path for the first step are determined. The obtained energy barrier of considered cases fall into the range of 0.02-0.43 eV, supporting the fact that the substituting B for C can easily realized even at room temperature. As the second step, after removing irrelevant groups in vertical direction through H saturation, the perfect B doping is realized at last. This work successfully explain the above experimental phenomenon and propose a novel and feasible method aiming at B doping of graphene.
• ### An Intelligent QoS Identification for Untrustworthy Web Services Via Two-phase Neural Networks(1602.07366)

Feb. 24, 2016 cs.SE
QoS identification for untrustworthy Web services is critical in QoS management in the service computing since the performance of untrustworthy Web services may result in QoS downgrade. The key issue is to intelligently learn the characteristics of trustworthy Web services from different QoS levels, then to identify the untrustworthy ones according to the characteristics of QoS metrics. As one of the intelligent identification approaches, deep neural network has emerged as a powerful technique in recent years. In this paper, we propose a novel two-phase neural network model to identify the untrustworthy Web services. In the first phase, Web services are collected from the published QoS dataset. Then, we design a feedforward neural network model to build the classifier for Web services with different QoS levels. In the second phase, we employ a probabilistic neural network (PNN) model to identify the untrustworthy Web services from each classification. The experimental results show the proposed approach has 90.5% identification ratio far higher than other competing approaches.
• ### HBTM: A Heartbeat-based Behavior Detection Mechanism for POSIX Threads and OpenMP Applications(1512.00665)

Dec. 2, 2015 cs.DC
Extreme-scale computing involves hundreds of millions of threads with multi-level parallelism running on large-scale hierarchical and heterogeneous hardware. In POSIX threads and OpenMP applications, some key behaviors occurring in runtime such as thread failure, busy waiting, and exit need to be accurately and timely detected. However, for the most of these applications, there are lack of unified and efficient detection mechanisms to do this. In this paper, a heartbeat-based behavior detection mechanism for POSIX threads (Pthreads) and OpenMP applications (HBTM) is proposed. In the design, two types of implementations are conducted, centralized and decentralized respectively. In both implementations, unified API has been designed to guarantee the generality of the mechanism. Meanwhile, a ring-based detection algorithm is designed to ease the burden of the centra thread at runtime. To evaluate the mechanism, the NAS Parallel Benchmarks (NPB) are used to test the performance of the HBTM. The experimental results show that the HBTM supports detection of behaviors of POSIX threads and OpenMP applications while acquiring a short latency and near 1% overhead.
• ### From Competition to Complementarity: Comparative Influence Diffusion and Maximization(1507.00317)

Nov. 5, 2015 physics.soc-ph, cs.SI
Influence maximization is a well-studied problem that asks for a small set of influential users from a social network, such that by targeting them as early adopters, the expected total adoption through influence cascades over the network is maximized. However, almost all prior work focuses on cascades of a single propagating entity or purely-competitive entities. In this work, we propose the Comparative Independent Cascade (Com-IC) model that covers the full spectrum of entity interactions from competition to complementarity. In Com-IC, users' adoption decisions depend not only on edge-level information propagation, but also on a node-level automaton whose behavior is governed by a set of model parameters, enabling our model to capture not only competition, but also complementarity, to any possible degree. We study two natural optimization problems, Self Influence Maximization and Complementary Influence Maximization, in a novel setting with complementary entities. Both problems are NP-hard, and we devise efficient and effective approximation algorithms via non-trivial techniques based on reverse-reachable sets and a novel "sandwich approximation". The applicability of both techniques extends beyond our model and problems. Our experiments show that the proposed algorithms consistently outperform intuitive baselines in four real-world social networks, often by a significant margin. In addition, we learn model parameters from real user action logs.