arXiv subjects

Xiang Yu

Publications and source records attributed to Xiang Yu.

At least 145 records · Page 8Linked to original sources

arXiv2021

Synthetic Generation of Three-Dimensional Cancer Cell Models from Histopathological Images

Synthetic generation of three-dimensional cell models from histopathological images aims to enhance understanding of cell mutation, and progression of cancer, necessary for clinical assessment and optimal treatment. Classical reconstruction algorithms based on image registration of consecutive slides of stained tissues are prone to errors and often not suitable for the training of three-dimensional segmentation algorithms. We propose a novel framework to generate synthetic three-dimensional histological models based on a generator-discriminator pattern optimizing constrained features that construct a 3D model via a Blender interface exploiting smooth shape continuity typical for biological specimens. To capture the spatial context of entire cell clusters we deploy a novel deep topology transformer that implements and attention mechanism on cell group images to extract features for the frozen feature decoder. The proposed algorithms achieves high quantitative and qualitative synthesis evident in comparative evaluation metrics such as a low Frechet-Inception scores.

eess.IV↗

arXiv2020

A refined dynamic finite-strain shell theory for incompressible hyperelastic materials: equations and two-dimensional shell virtual work principle

Based on previous work for the static problem, in this paper we first derive one form of dynamic finite-strain shell equations for incompressible hyperelastic materials that involve three shell constitutive relations. In order to single out the bending effect as well as to reduce the number of shell constitutive relations, a further refinement is performed, which leads to a refined dynamic finite-strain shell theory with only two shell constitutive relations (deducible from the given three-dimensional (3D) strain energy function) and some new insights are also deduced. By using the weak formulation of the shell equations and the variation of the 3D Lagrange functional, boundary conditions and the two-dimensional (2D) shell virtual work principle are derived. As a benchmark problem, we consider the extension and inflation of an arterial segment. The good agreement between the asymptotic solution based on the shell equations and that from the 3D exact one gives verification of the former. The refined shell theory is also applied to study the plane-strain vibrations of a pressurized artery, and the effects of the axial pre-stretch, pressure and fibre angle on the vibration frequencies are investigated in detail.

cs.CE↗

arXiv2020

Uncertainty-Aware Physically-Guided Proxy Tasks for Unseen Domain Face Anti-spoofing

Face anti-spoofing (FAS) seeks to discriminate genuine faces from fake ones arising from any type of spoofing attack. Due to the wide varieties of attacks, it is implausible to obtain training data that spans all attack types. We propose to leverage physical cues to attain better generalization on unseen domains. As a specific demonstration, we use physically guided proxy cues such as depth, reflection, and material to complement our main anti-spoofing (a.k.a liveness detection) task, with the intuition that genuine faces across domains have consistent face-like geometry, minimal reflection, and skin material. We introduce a novel uncertainty-aware attention scheme that independently learns to weigh the relative contributions of the main and proxy tasks, preventing the over-confident issue with traditional attention modules. Further, we propose attribute-assisted hard negative mining to disentangle liveness-irrelevant features with liveness features during learning. We evaluate extensively on public benchmarks with intra-dataset and inter-dataset protocols. Our method achieves the superior performance especially in unseen domain generalization for FAS.

cs.CV↗

arXiv2020

WeiPS: a symmetric fusion model framework for large-scale online learning

The recommendation system is an important commercial application of machine learning, where billions of feed views in the information flow every day. In reality, the interaction between user and item usually makes user's interest changing over time, thus many companies (e.g. ByteDance, Baidu, Alibaba, and Weibo) employ online learning as an effective way to quickly capture user interests. However, hundreds of billions of model parameters present online learning with challenges for real-time model deployment. Besides, model stability is another key point for online learning. To this end, we design and implement a symmetric fusion online learning system framework called WeiPS, which integrates model training and model inference. Specifically, WeiPS carries out second level model deployment by streaming update mechanism to satisfy the consistency requirement. Moreover, it uses multi-level fault tolerance and real-time domino degradation to achieve high availability requirement.

cs.LG↗

arXiv2020

Optimal Dividend Strategy for an Insurance Group with Contagious Default Risk

This paper studies the optimal dividend for a multi-line insurance group, in which each subsidiary runs a product line and is exposed to some external credit risk. The default contagion is considered such that one default event may increase the default probabilities of all surviving subsidiaries. The total dividend problem for the insurance group is investigated and we find that the optimal dividend strategy is still of the barrier type. Furthermore, we show that the optimal barrier of each subsidiary is modulated by the default state. That is, how many and which subsidiaries have defaulted will determine the dividend threshold of each surviving subsidiary. These conclusions are based on the analysis of the associated recursive system of Hamilton-Jacobi-Bellman variational inequalities (HJBVIs). The existence of the classical solution is established and the verification theorem is proved. In the case of two subsidiaries, the value function and optimal barriers are given in analytical forms, allowing us to conclude that the optimal barrier of one subsidiary decreases if the other subsidiary defaults.

q-fin.RM↗

arXiv2020

Lifetime Ruin under High-watermark Fees and Drift Uncertainty

This paper aims to make a new contribution to the study of lifetime ruin problem by considering investment in two hedge funds with high-watermark fees and drift uncertainty. Due to multi-dimensional performance fees that are charged whenever each fund profit exceeds its historical maximum, the value function is expected to be multi-dimensional. New mathematical challenges arise as the standard dimension reduction cannot be applied, and the convexity of the value function and Isaacs condition may not hold in our ruin probability minimization problem with drift uncertainty. We propose to employ the stochastic Perron's method to characterize the value function as the unique viscosity solution to the associated Hamilton Jacobi Bellman (HJB) equation without resorting to the proof of dynamic programming principle. The required comparison principle is also established in our setting to close the loop of stochastic Perron's method.

q-fin.MF↗

arXiv2020

Improving Face Recognition by Clustering Unlabeled Faces in the Wild

While deep face recognition has benefited significantly from large-scale labeled data, current research is focused on leveraging unlabeled data to further boost performance, reducing the cost of human annotation. Prior work has mostly been in controlled settings, where the labeled and unlabeled data sets have no overlapping identities by construction. This is not realistic in large-scale face recognition, where one must contend with such overlaps, the frequency of which increases with the volume of data. Ignoring identity overlap leads to significant labeling noise, as data from the same identity is split into multiple clusters. To address this, we propose a novel identity separation method based on extreme value theory. It is formulated as an out-of-distribution detection algorithm, and greatly reduces the problems caused by overlapping-identity label noise. Considering cluster assignments as pseudo-labels, we must also overcome the labeling noise from clustering errors. We propose a modulation of the cosine loss, where the modulation weights correspond to an estimate of clustering uncertainty. Extensive experiments on both controlled and real settings demonstrate our method's consistent improvements over supervised baselines, e.g., 11.6% improvement on IJB-A verification.

cs.CV↗

arXiv2020

Distinct coordinate solutions of linear equations over finite fields

Let $\mathbb{F}_q$ be the finite field of $q$ elements and $a_1,a_2, \ldots, a_k, b\in \mathbb{F}_q$. We investigate $N_{\mathbb{F}_q}(a_1, a_2, \ldots,a_k;b)$, the number of ordered solutions $(x_1, x_2, \ldots,x_k)\in\mathbb{F}_q^k$ of the linear equation $$ a_1x_1+a_2x_2+\cdots+a_kx_k=b$$ with all $x_i$ distinct. We obtain an explicit formula for $N_{\mathbb{F}_q}(a_1,a_2, \ldots, a_k;b)$ involving combinatorial numbers depending on $a_i$'s. In particular, we obtain closed formulas for two special cases. One is that $a_i, 1\leq i\leq k$ take at most three distinct values and the other is that $\sum_{i=1}^ka_i=0$ and $\sum_{i\in I}a_i\neq 0$ for any $I\subsetneq [k]$. The same technique works when $\mathbb{F}_q$ is replaced by $\mathbb{Z}_n$, the ring of integers modulo $n$. In particular, we give a new proof for the main result given by Bibak, Kapron and Srinivasan, which generalizes a theorem of Schönemann via a graph theoretic method.

math.NT↗

arXiv2020

On Periodic orbits of the Planar N-body Problem

By introducing a new coordinate system, we prove that there are abundant new periodic orbits near relative equilibrium solutions of the N-body problem. We consider only Lagrange relative equilibrium of the three-body problem and Euler-Moulton relative equilibrium of the N-body problem, although we believe that there are similar results for general relative equilibrium solutions. All of these periodic orbits lie on a 2d-dimensional central manifold of the planar N-body problem. Besides d one parameter family of periodic orbits which are well known as Lyapunov's orbits or Weinstein's orbits, we further prove that periodic orbits are unexpectedly abundant: generically the relative measure of the closure of the set of periodic orbits near relative equilibrium solutions on the 2d-dimensional central manifold is close to 1. These abundant periodic orbits are named Conley-Zender's orbits, since to find them is based on an extended result of Conley and Zender on the local existence result for periodic orbits near an elliptic equilibrium point of a Hamiltonian. In particular, the results provide some evidences to support the well known claim of Poincaré on the conjecture of periodic orbits of the N-body problem.

math.DS↗

arXiv2020

On Dynamic Programming Principle for Stochastic Control under Expectation Constraints

This paper studies the dynamic programming principle using the measurable selection method for stochastic control of continuous processes. The novelty of this work is to incorporate intermediate expectation constraints on the canonical space at each time t. Motivated by some financial applications, we show that several types of dynamic trading constraints can be reformulated into expectation constraints on paths of controlled state processes. Our results can therefore be employed to recover the dynamic programming principle for these optimal investment problems under dynamic constraints, possibly path-dependent, in a non-Markovian framework.

math.OC↗

arXiv2020

Towards Universal Representation Learning for Deep Face Recognition

Recognizing wild faces is extremely hard as they appear with all kinds of variations. Traditional methods either train with specifically annotated variation data from target domains, or by introducing unlabeled target variation data to adapt from the training data. Instead, we propose a universal representation learning framework that can deal with larger variation unseen in the given training data without leveraging target domain knowledge. We firstly synthesize training data alongside some semantically meaningful variations, such as low resolution, occlusion and head pose. However, directly feeding the augmented data for training will not converge well as the newly introduced samples are mostly hard examples. We propose to split the feature embedding into multiple sub-embeddings, and associate different confidence values for each sub-embedding to smooth the training procedure. The sub-embeddings are further decorrelated by regularizing variation classification loss and variation adversarial loss on different partitions of them. Experiments show that our method achieves top performance on general face recognition datasets such as LFW and MegaFace, while significantly better on extreme benchmarks such as TinyFace and IJB-S.

cs.CV↗

arXiv2020

On the bail-out dividend problem for spectrally negative Markov additive models

This paper studies the bail-out optimal dividend problem with regime switching under the constraint that the cumulative dividend strategy is absolutely continuous. We confirm the optimality of the regime-modulated refraction-reflection strategy when the underlying risk model follows a general spectrally negative Markov additive process. To verify the conjecture of a barrier type optimal control, we first introduce and study an auxiliary problem with the final payoff at an exponential terminal time and characterize the optimal threshold explicitly using fluctuation identities of the refracted-reflected Levy process. Second, we transform the problem with regime-switching into an equivalent local optimization problem with a final payoff up to the first regime switching time. The refraction-reflection strategy with regime-modulated thresholds can be shown as optimal by using results in the first step and some fixed point arguments for auxiliary recursive iterations.

q-fin.MF↗

arXiv2019

DAVID: Dual-Attentional Video Deblurring

Blind video deblurring restores sharp frames from a blurry sequence without any prior. It is a challenging task because the blur due to camera shake, object movement and defocusing is heterogeneous in both temporal and spatial dimensions. Traditional methods train on datasets synthesized with a single level of blur, and thus do not generalize well across levels of blurriness. To address this challenge, we propose a dual attention mechanism to dynamically aggregate temporal cues for deblurring with an end-to-end trainable network structure. Specifically, an internal attention module adaptively selects the optimal temporal scales for restoring the sharp center frame. An external attention module adaptively aggregates and refines multiple sharp frame estimates, from several internal attention modules designed for different blur levels. To train and evaluate on more diverse blur severity levels, we propose a Challenging DVD dataset generated from the raw DVD video set by pooling frames with different temporal windows. Our framework achieves consistently better performance on this more challenging dataset while obtaining strongly competitive results on the original DVD benchmark. Extensive ablative studies and qualitative visualizations further demonstrate the advantage of our method in handling real video blur.

cs.CV↗

arXiv2019

Regular polygonal equilibrium configurations on S^1 and stability of the associated relative equilibria

For the curved n-body problem in S^3, we show that a regular polygonal configuration for n masses on a geodesic is an equilibrium configuration if and only if n is odd and the masses are equal. The equilibrium configuration is associated with a one-parameter family (depending on the angular velocity) of relative equilibria, which take place on S^1 embedded in S^2. We then study the stability of the associated relative equilibria on two invariant manifolds, T^*((§^1)^n\D) and T^*((§^2)^n\D). We show that they are Lyapunov stable on S^1, they are Lyapunov stable on S^2 if the absolute value of angular velocity is larger than a certain value, and that they are linearly unstable on S^2 if the absolute value of angular velocity is smaller than that certain value.

math.DS↗

arXiv2019

Feature Transfer Learning for Deep Face Recognition with Under-Represented Data

Despite the large volume of face recognition datasets, there is a significant portion of subjects, of which the samples are insufficient and thus under-represented. Ignoring such significant portion results in insufficient training data. Training with under-represented data leads to biased classifiers in conventionally-trained deep networks. In this paper, we propose a center-based feature transfer framework to augment the feature space of under-represented subjects from the regular subjects that have sufficiently diverse samples. A Gaussian prior of the variance is assumed across all subjects and the variance from regular ones are transferred to the under-represented ones. This encourages the under-represented distribution to be closer to the regular distribution. Further, an alternating training regimen is proposed to simultaneously achieve less biased classifiers and a more discriminative feature representation. We conduct ablative study to mimic the under-represented datasets by varying the portion of under-represented classes on the MS-Celeb-1M dataset. Advantageous results on LFW, IJB-A and MS-Celeb-1M demonstrate the effectiveness of our feature transfer and training strategy, compared to both general baselines and state-of-the-art methods. Moreover, our feature transfer successfully presents smooth visual interpolation, which conducts disentanglement to preserve identity of a class while augmenting its feature space with non-identity variations such as pose and lighting.

cs.CV↗

arXiv2019

Utility maximization with proportional transaction costs under model uncertainty

We consider a discrete time financial market with proportional transaction costs under model uncertainty, and study a numéraire-based semi-static utility maximization problem with an exponential utility preference. The randomization techniques recently developed in \cite{BDT17} allow us to transform the original problem into a frictionless counterpart on an enlarged space. By suggesting a different dynamic programming argument than in \cite{bartl2016exponential}, we are able to prove the existence of the optimal strategy and the convex duality theorem in our context with transaction costs. In the frictionless framework, this alternative dynamic programming argument also allows us to generalize the main results in \cite{bartl2016exponential} to a weaker market condition. Moreover, as an application of the duality representation, some basic features of utility indifference prices are investigated in our robust setting with transaction costs.

q-fin.MF↗

arXiv2019

Pose-variant 3D Facial Attribute Generation

We address the challenging problem of generating facial attributes using a single image in an unconstrained pose. In contrast to prior works that largely consider generation on 2D near-frontal images, we propose a GAN-based framework to generate attributes directly on a dense 3D representation given by UV texture and position maps, resulting in photorealistic, geometrically-consistent and identity-preserving outputs. Starting from a self-occluded UV texture map obtained by applying an off-the-shelf 3D reconstruction method, we propose two novel components. First, a texture completion generative adversarial network (TC-GAN) completes the partial UV texture map. Second, a 3D attribute generation GAN (3DA-GAN) synthesizes the target attribute while obtaining an appearance consistent with 3D face geometry and preserving identity. Extensive experiments on CelebA, LFW and IJB-A show that our method achieves consistently better attribute generation accuracy than prior methods, a higher degree of qualitative photorealism and preserves face identity information.

cs.CV↗

arXiv2019

New homoclinic orbits for Hamiltonian systems with asymptotically quadratic growth at infinity

In this paper, we study the existence and multiplicity of homoclinic solutions for following Hamiltonian systems with asymptotically quadratic nonlinearities at infinity \begin{eqnarray*} \ddot{u}(t)-L(t)u+\nabla W(t,u)=0. {eqnarray*} We introduce a new coercive condition and obtain a new embedding theorem. With this theorem, we show that above systems possess at least one nontrivial homoclinic orbits by Generalized Mountain Pass Theorem. By Variant Fountain Theorem, infinitely many homoclinic orbits are obtained for above problem with symmetric condition. Our asymptotically quadratic conditions are different from previous ones in the references.

math.DS↗