arXiv subjects

Yifan Chen

Publications and source records attributed to Yifan Chen.

At least 235 records · Page 13Linked to original sources

arXiv2018

Nanorobots-assisted Detection of Multifocal Cancer: A Multimodal Optimization Perspective

We propose a new framework of computing-inspired multifocal cancer detection procedure (MCDP). Under the rubric of MCDP, the tumor foci to be detected are regarded as solutions of the objective function, the tissue region around the cancer areas represents the parameter space, and the nanorobots loaded with contrast medium molecules for cancer detection correspond to the optimization agents. The process that the nanorobots detect tumors by swimming in the high-risk tissue region can be regarded as the process that the agents search for the solutions of an objective function in the parameter space with some constraints. For multimodal optimization (MMO) aiming to locate multiple optimal solutions in a single simulation run, niche technology has been widely used. Specifically, the Niche Genetic Algorithm (NGA) has been shown to be particularly effective in solving MMO. It can be used to identify the global optima of multiple hump functions in a running, keep effectively the diversity of the population, and avoid premature of the genetic algorithm. Learning from the optimization procedure of NGA, we propose the NGA-inspired MCDP in order to locate the tumor targets efficiently while taking into account realistic in vivo propagation and controlling of nanorobots, which is different from the use scenario of the standard NGA. To improve the performance of the MCDP, we also modify the crossover operator of the original NGA from crossing within a population to crossing between two populations. Finally, we present comprehensive numerical examples to demonstrate the effectiveness of the NGA-inspired MCDP when the biological objective function is associated with the blood flow velocity profile caused by tumor-induced angiogenesis.

q-bio.TO↗

arXiv2018

Biosensing-by-learning Direct Targeting Strategy for Enhanced Tumor Sensitization

Objective:We propose a novel iterative-optimizationinspired direct targeting strategy (DTS) for smart nanosystems, which harness swarms of externally manipulable nanoswimmers assembled by magnetic nanoparticles (MNPs) for knowledgeaided tumor sensitization and targeting. We aim to demonstrate through computational experiments that the proposed DTS can significantly enhance the accumulation of MNPs in the tumor site, which serve as a contrast agent in various medical imaging modalities, by using the shortest possible physiological routes and with minimal systemic exposure. Methods: The epicenter of a tumor corresponds to the global maximum of an externally measurable objective function associated with an in vivo tumor-triggered biophysical gradient; the domain of the objective function is the tissue region at a high risk of malignancy; swarms of externally controllable magnetic nanoswimmers for tumor sensitization are modeled as the guess inputs. The objective function may be resulted from a passive phenomenon such as reduced blood flow or increased kurtosis of microvasculature due to tumor angiogenesis; otherwise, the objective function may involve an active phenomenon such as the fibrin formed during the coagulation cascade activated by tumortargeted "activator" nanoparticles. Subsequently, the DTS can be interpreted from the iterative optimization perspective: guess inputs (i.e., swarms of nanoswimmers) are continuously updated according to the gradient of the objective function in order to find the optimum (i.e., tumor) by moving through the domain (i.e., tissue under screening). Along this line of thought, we propose the computational model based on the gradient descent (GD) iterative method to describe the GD-inspired DTS, which takes into account the realistic in vivo propagation scenario of nanoswimmers.

physics.med-ph↗

arXiv2018

Regularized Ensembles and Transferability in Adversarial Learning

Despite the considerable success of convolutional neural networks in a broad array of domains, recent research has shown these to be vulnerable to small adversarial perturbations, commonly known as adversarial examples. Moreover, such examples have shown to be remarkably portable, or transferable, from one model to another, enabling highly successful black-box attacks. We explore this issue of transferability and robustness from two dimensions: first, considering the impact of conventional $l_p$ regularization as well as replacing the top layer with a linear support vector machine (SVM), and second, the value of combining regularized models into an ensemble. We show that models trained with different regularizers present barriers to transferability, as does partial information about the models comprising the ensemble.

cs.LG↗

arXiv2018

Prostate Segmentation using 2D Bridged U-net

In this paper, we focus on three problems in deep learning based medical image segmentation. Firstly, U-net, as a popular model for medical image segmentation, is difficult to train when convolutional layers increase even though a deeper network usually has a better generalization ability because of more learnable parameters. Secondly, the exponential ReLU (ELU), as an alternative of ReLU, is not much different from ReLU when the network of interest gets deep. Thirdly, the Dice loss, as one of the pervasive loss functions for medical image segmentation, is not effective when the prediction is close to ground truth and will cause oscillation during training. To address the aforementioned three problems, we propose and validate a deeper network that can fit medical image datasets that are usually small in the sample size. Meanwhile, we propose a new loss function to accelerate the learning process and a combination of different activation functions to improve the network performance. Our experimental results suggest that our network is comparable or superior to state-of-the-art methods.

cs.CV↗

arXiv2018

Perturbative moduli stabilisation in type IIB/F-theory framework

We propose a new mechanism of (geometric) moduli stabilisation in type IIB/F-theory four-dimensional compactifications on Calabi-Yau manifolds, in the presence of 7-branes, that does not rely on non-perturbative effects. Complex structure moduli and the axion-dilaton system are stabilised in the standard way, without breaking supersymmetry, using 3-form internal fluxes. Kähler class moduli stabilisation utilises perturbative string loop corrections, together with internal magnetic fields along the D7-branes world-volume leading to Fayet-Iliopoulos D-terms in the effective supergravity action. The main ingredient that makes the stabilisation possible at a de Sitter vacuum is the logarithmic dependence of the string loop corrections in the large two-dimensional transverse volume limit of the 7-branes.

hep-th↗

arXiv2018

Minimal constrained superfields and the Fayet-Iliopoulos model

We show how the necessary constraints to project out all the components of a chiral superfield except for some scalar degrees of freedom originate from simple operators in the microscopic theory. This is in particular useful in constructing the simplest models of a goldstone boson/inflaton; or extracting the Standard Model Higgs doublet from a supersymmetric electroweak sector. We use the Fayet-Iliopoulos model as an example of the origin for the supersymmetry breaking. We consider the regime where both gauge symmetry and supersymmetry are spontaneously broken, leaving (in the decoupling limit) the goldstino as the only light mode in this sector. We show in three different ways, both in components and in superspace language, how the nilpotent goldstino superfield emerges. We then use it to write different effective operators and extract some of the consequences for the low energy spectrum.

hep-th↗

arXiv2018

A post-processing method to improve the white matter hyperintensity segmentation accuracy for randomly-initialized U-net

White matter hyperintensity (WMH) is commonly found in elder individuals and appears to be associated with brain diseases. U-net is a convolutional network that has been widely used for biomedical image segmentation. Recently, U-net has been successfully applied to WMH segmentation. Random initialization is usally used to initialize the model weights in the U-net. However, the model may coverage to different local optima with different randomly initialized weights. We find a combination of thresholding and averaging the outputs of U-nets with different random initializations can largely improve the WMH segmentation accuracy. Based on this observation, we propose a post-processing technique concerning the way how averaging and thresholding are conducted. Specifically, we first transfer the score maps from three U-nets to binary masks via thresholding and then average those binary masks to obtain the final WMH segmentation. Both quantitative analysis (via the Dice similarity coefficient) and qualitative analysis (via visual examinations) reveal the superior performance of the proposed method. This post-processing technique is independent of the model used. As such, it can also be applied to situations where other deep learning models are employed, especially when random initialization is adopted and pre-training is unavailable.

cs.CV↗

arXiv2018

A Collective Variational Autoencoder for Top-$N$ Recommendation with Side Information

Recommender systems have been studied extensively due to their practical use in many real-world scenarios. Despite this, generating effective recommendations with sparse user ratings remains a challenge. Side information associated with items has been widely utilized to address rating sparsity. Existing recommendation models that use side information are linear and, hence, have restricted expressiveness. Deep learning has been used to capture non-linearities by learning deep item representations from side information but as side information is high-dimensional existing deep models tend to have large input dimensionality, which dominates their overall size. This makes them difficult to train, especially with small numbers of inputs. Rather than learning item representations, which is problematic with high-dimensional side information, in this paper, we propose to learn feature representation through deep learning from side information. Learning feature representations, on the other hand, ensures a sufficient number of inputs to train a deep network. To achieve this, we propose to simultaneously recover user ratings and side information, by using a Variational Autoencoder (VAE). Specifically, user ratings and side information are encoded and decoded collectively through the same inference network and generation network. This is possible as both user ratings and side information are data associated with items. To account for the heterogeneity of user rating and side information, the final layer of the generation network follows different distributions depending on the type of information. The proposed model is easy to implement and efficient to optimize and is shown to outperform state-of-the-art top-$N$ recommendation methods that use side information.

cs.IR↗

arXiv2018

Run-and-Inspect Method for Nonconvex Optimization and Global Optimality Bounds for R-Local Minimizers

Many optimization algorithms converge to stationary points. When the underlying problem is nonconvex, they may get trapped at local minimizers and occasionally stagnate near saddle points. We propose the Run-and-Inspect Method, which adds an "inspect" phase to existing algorithms that helps escape from non-global stationary points. The inspection samples a set of points in a radius $R$ around the current point. When a sample point yields a sufficient decrease in the objective, we move there and resume an existing algorithm. If no sufficient decrease is found, the current point is called an approximate $R$-local minimizer. We show that an $R$-local minimizer is globally optimal, up to a specific error depending on $R$, if the objective function can be implicitly decomposed into a smooth convex function plus a restricted function that is possibly nonconvex, nonsmooth. For high-dimensional problems, we introduce blockwise inspections to overcome the curse of dimensionality while still maintaining optimality bounds up to a factor equal to the number of blocks. Our method performs well on a set of artificial and realistic nonconvex problems by coupling with gradient descent, coordinate descent, EM, and prox-linear algorithms.

math.OC↗

arXiv2018

Stripping a supermultiplet of all but one scalar: a microscopic view

We explain how a single scalar degree of freedom can be obtained from projecting out all the other components of a chiral superfield using simple operators in the microscopic theory. We use the Fayet-Iliopoulos model as an example of the origin for the necessary supersymmetry breaking. We also comment on one peculiar aspect where non-linear realisation can be useful: the freeze-in scenario for gravitino dark matter.

hep-th↗

arXiv2018

Molecular Communications at the Macroscale: A Novel Framework for Modeling Epidemic Spreading and Mitigation

Using the notion of effective distance proposed by Brockmann and Helbing, complex spatiotemporal processes of epidemic spreading can be reduced to circular wave propagation patterns with well-defined wavefronts. This hidden homogeneity of contagion phenomena enables the mapping of virtual mobility networks to physical propagation channels. Subsequently, we propose a novel communications-inspired model of epidemic spreading and mitigation by establishing the one-to-one correspondence between the essential components comprising information and disease transmissions. The epidemic processes can be regarded as macroscale molecular communications, in which individuals are macroscale information molecules carrying messages (epidemiological states). We then present the notions of normalized ensemble-average prevalence (NEAP) and prevalence delay profile (PDP) to characterize the relative impact and time difference of all the spreading paths, which are analogous to the classical description methods of path loss and power delay profile in communications. Furthermore, we introduce the metric of root mean square (RMS) delay spread to measure the distortion of early contagion dynamics caused by multiple infection transmission routes. In addition, we show how social and medical interventions can be understood from the perspectives of various communication modules. The proposed framework provides an intuitive, coherent, and efficient approach for characterization of the disease outbreaks by applying the deep-rooted communications theories as the analytical lens.

physics.soc-ph↗

arXiv2017

The quadratic Wasserstein metric for earthquake location

In [Engquist et al., Commun. Math. Sci., 14(2016)], the Wasserstein metric was successfully introduced to the full waveform inversion. We apply this method to the earthquake location problem. For this problem, the seismic stations are far from each other. Thus, the trace by trace comparison [Yang et al., arXiv(2016)] is a natural way to compare the earthquake signals. Under this framework, we have derived a concise analytic expression of the Frèchet gradient of the Wasserstein metric, which leads to a simple and efficient implementation for the adjoint method. We square and normalize the earthquake signals for comparison so that the convexity of the misfit function with respect to earthquake hypocenter and origin time can be observed numerically. To reduce the impact of noise, which can not offset each other after squaring the signals, a new control parameter is introduced. Finally, the LMF (Levenberg-Marquardt-Fletcher) method is applied to solve the resulted optimization problem. According to the numerical experiments, only a few iterations are required to converge to the real earthquake hypocenter and origin time. Even for data with noise, we can obtain reasonable and convergent numerical results.

math.NA↗

arXiv2017

A Characterization of Inoue Surfaces with $p_g=0$ and $K^2=7$

Inoue constructed the first examples of smooth minimal complex surfaces of general type with $p_g=0$ and $K^2=7$.These surfaces are finite Galois covers of the $4$-nodal cubic surface with the Galois group, the Klein group $\mathbb{Z}_2\times \mathbb{Z}_2$. For such a surface $S$, the bicanonical map of $S$ has degree $2$ and it is composed with exactly one involution in the Galois group. The divisorial part of the fixed locus of this involution consists of two irreducible components:one is a genus $3$ curve with self-intersection number $0$ and the other is a genus $2$ curve with self-intersection number $-1$. Conversely, assume that $S$ is a smooth minimal complex surface of general type with $p_g=0$, $K^2=7$ and having an involution $σ$. We show that, if the divisorial part of the fixed locus of $σ$ consists of two irreducible components $R_1$ and $R_2$,with $g(R_1)=3, R_1^2=0, g(R_2)=2$ and $R_2^2=-1$, then the Klein group $\mathbb{Z}_2\times \mathbb{Z}_2$ acts faithfully on $S$ and $S$ is indeed an Inoue surface.

math.AG↗

arXiv2017

Leveraging High-Dimensional Side Information for Top-N Recommendation

Top-$N$ recommender systems typically utilize side information to address the problem of data sparsity. As nowadays side information is growing towards high dimensionality, the performances of existing methods deteriorate in terms of both effectiveness and efficiency, which imposes a severe technical challenge. In order to take advantage of high-dimensional side information, we propose in this paper an embedded feature selection method to facilitate top-$N$ recommendation. In particular, we propose to learn feature weights of side information, where zero-valued features are naturally filtered out. We also introduce non-negativity and sparsity to the feature weights, to facilitate feature selection and encourage low-rank structure. Two optimization problems are accordingly put forward, respectively, where the feature selection is tightly or loosely coupled with the learning procedure. Augmented Lagrange Multiplier and Alternating Direction Method are applied to efficiently solve the problems. Experiment results demonstrate the superior recommendation quality of the proposed algorithm to that of the state-of-the-art alternatives.

cs.IR↗

arXiv2017

Spoof Surface Plasmon Polariton Leaky-Wave Antennas using Periodically Loaded Patches above PEC and AMC Ground Planes

This paper proposes two spoof surface plasmon polariton (SSPP) leaky-wave antennas using periodically loaded patches above perfect electric conductor (PEC) and artificial magnetic conductor (AMC) ground planes, respectively. The SSPP leaky-wave antenna is based on a SSPP transmission line, along which circular patches are periodically loaded on both sides to provide an additional momentum for phase matching with the radiated waves in the air. The PEC and AMC ground planes underneath the antenna reflect the radiated waves into the upward space, leading to an enhanced radiation gain. Both PEC- and AMC-grounded antenna prototypes are fabricated and measured in comparison with the one without any ground plane. The experimental results show that the PEC and AMC ground planes increase the radiation gain by approximately 3 dB within the operational frequency range 4.5-6.5 GHz. It also demonstrates that the AMC-grounded leaky-wave antenna, with a thickness of 0.08λat 6 GHz, features more compact profile than the PEC-grounded one (with a thickness of 0.3λat 6 GHz).

physics.app-ph↗

arXiv2017

A Minimal Model of Gravitino Dark Matter

Motivated by the absence of signals of new physics in both searches for new particles at LHC and for a Weakly Interacting Massive Particle (WIMP) dark matter candidate, we consider a scenario where supersymmetry is broken at a scale above the reheating temperature. The low energy particle content consists then only in Standard Model states and a gravitino. We investigate the possibility that the latter provides the main component of dark matter through a freeze in mechanism from annihilation of thermalized Standard Model particles. We focus on the case where its production through scattering in the thermal plasma is well approximated by the non-linear supersymmetric effective Lagrangian of the associated goldstino and identify the parameter space allowed by the cosmological constraints, allowing the possibility of large reheating temperature compatible with leptogenesis scenarios, alleviating the so called "gravitino problem".

hep-ph↗

arXiv2017

Capacitor-Loaded Spoof Surface Plasmon (SSP) for Flexible Dispersion Control and High-Selectivity Filtering

This letter proposes a new spoof surface plasmon transmission line (SSP-TL) using capacitor loading techniques. This new SSP-TL features flexible and reconfigurable dispersion control and highly selective filtering performance without resorting to configuration change. Moreover, it requires much smaller line width than the conventional SSP-TLs for achieving a extremely slow wave (or a highly confined field), which is quite useful for a compact system. To illustrate the design principle, several examples are designed within the frequency range of 2-8 GHz. Both numerical and experimental results are given in comparison with the conventional SSP-TL. It is demonstrated that the proposed technique provides a better performance in size reduction and dispersion reconfigurability.

physics.class-ph↗

arXiv2016

Content-Based Top-N Recommendation using Heterogeneous Relations

Top-$N$ recommender systems have been extensively studied. However, the sparsity of user-item activities has not been well resolved. While many hybrid systems were proposed to address the cold-start problem, the profile information has not been sufficiently leveraged. Furthermore, the heterogeneity of profiles between users and items intensifies the challenge. In this paper, we propose a content-based top-$N$ recommender system by learning the global term weights in profiles. To achieve this, we bring in PathSim, which could well measures the node similarity with heterogeneous relations (between users and items). Starting from the original TF-IDF value, the global term weights gradually converge, and eventually reflect both profile and activity information. To facilitate training, the derivative is reformulated into matrix form, which could easily be paralleled. We conduct extensive experiments, which demonstrate the superiority of the proposed method.

cs.IR↗