Search arXiv⌕ Search

arXiv subjects

Kai Zhang

Publications and source records attributed to Kai Zhang.

At least 613 records · Page 34Linked to original sources
arXiv2020

How Does Gender Balance In Training Data Affect Face Recognition Accuracy?

Deep learning methods have greatly increased the accuracy of face recognition, but an old problem still persists: accuracy is usually higher for men than women. It is often speculated that lower accuracy for women is caused by under-representation in the training data. This work investigates female under-representation in the training data is truly the cause of lower accuracy for females on test data. Using a state-of-the-art deep CNN, three different loss functions, and two training datasets, we train each on seven subsets with different male/female ratios, totaling forty two trainings, that are tested on three different datasets. Results show that (1) gender balance in the training data does not translate into gender balance in the test accuracy, (2) the "gender gap" in test accuracy is not minimized by a gender-balanced training set, but by a training set with more male images than female images, and (3) training to minimize the accuracy gap does not result in highest female, male or average accuracy

cs.CV↗
arXiv2020

Computational Performance of a Germline Variant Calling Pipeline for Next Generation Sequencing

With the booming of next generation sequencing technology and its implementation in clinical practice and life science research, the need for faster and more efficient data analysis methods becomes pressing in the field of sequencing. Here we report on the evaluation of an optimized germline mutation calling pipeline, HummingBird, by assessing its performance against the widely accepted BWA-GATK pipeline. We found that the HummingBird pipeline can significantly reduce the running time of the primary data analysis for whole genome sequencing and whole exome sequencing while without significantly sacrificing the variant calling accuracy. Thus, we conclude that expansion of such software usage will help to improve the primary data analysis efficiency for next generation sequencing.

q-bio.GN↗
arXiv2020

Deep Unfolding Network for Image Super-Resolution

Learning-based single image super-resolution (SISR) methods are continuously showing superior effectiveness and efficiency over traditional model-based methods, largely due to the end-to-end training. However, different from model-based methods that can handle the SISR problem with different scale factors, blur kernels and noise levels under a unified MAP (maximum a posteriori) framework, learning-based methods generally lack such flexibility. To address this issue, this paper proposes an end-to-end trainable unfolding network which leverages both learning-based methods and model-based methods. Specifically, by unfolding the MAP inference via a half-quadratic splitting algorithm, a fixed number of iterations consisting of alternately solving a data subproblem and a prior subproblem can be obtained. The two subproblems then can be solved with neural modules, resulting in an end-to-end trainable, iterative network. As a result, the proposed network inherits the flexibility of model-based methods to super-resolve blurry, noisy images for different scale factors via a single model, while maintaining the advantages of learning-based methods. Extensive experiments demonstrate the superiority of the proposed deep unfolding network in terms of flexibility, effectiveness and also generalizability.

eess.IV↗
arXiv2020

Neural Blind Deconvolution Using Deep Priors

Blind deconvolution is a classical yet challenging low-level vision problem with many real-world applications. Traditional maximum a posterior (MAP) based methods rely heavily on fixed and handcrafted priors that certainly are insufficient in characterizing clean images and blur kernels, and usually adopt specially designed alternating minimization to avoid trivial solution. In contrast, existing deep motion deblurring networks learn from massive training images the mapping to clean image or blur kernel, but are limited in handling various complex and large size blur kernels. To connect MAP and deep models, we in this paper present two generative networks for respectively modeling the deep priors of clean image and blur kernel, and propose an unconstrained neural optimization solution to blind deconvolution. In particular, we adopt an asymmetric Autoencoder with skip connections for generating latent clean image, and a fully-connected network (FCN) for generating blur kernel. Moreover, the SoftMax nonlinearity is applied to the output layer of FCN to meet the non-negative and equality constraints. The process of neural optimization can be explained as a kind of "zero-shot" self-supervised learning of the generative networks, and thus our proposed method is dubbed SelfDeblur. Experimental results show that our SelfDeblur can achieve notable quantitative gains as well as more visually plausible deblurring results in comparison to state-of-the-art blind deconvolution methods on benchmark datasets and real-world blurry images. The source code is available at https://github.com/csdwren/SelfDeblur

cs.CV↗
arXiv2020

Magnetization Current Simulation of High Temperature Bulk Superconductors Using A-V-A Formulation and Iterative Algorithm Method: Critical State Model and Flux Creep Model

In this work we will introduce the A-V-A formulation based iterative algorithm method (IAM) for simulating the magnetization current of high temperature superconductors. This new method embedded in ANSYS can simulate the critical state model by forcing the trapped current density to the critical current density Jc for all meshed superconducting elements after each iterative load step, as well as simulate the flux creep model by updating the E-J power law based resistivity values. The simulation results of a disk-shaped ReBCO bulk during zero field cooling (ZFC) or field cooling (FC) magnetization agree well with the simulation results from using the H-formulation in COMSOL. The computation time is shortened by using the A-V formulation in superconductor areas and the A-formulation in non-superconductor areas. This iterative method is further proved friendly for adding ferromagnetic materials into the FEA model or taking into account the magnetic field-dependent or mechanical strain-related critical current density of the superconductors. The influence factors for the magnetization simulation, including the specified iterative load steps, the initial resistivity, the ramping time and the updating coefficient, are discussed in detail. The A-V-A formulation based IAM, implemented in ANSYS, shows its unique advantages in adjustable computation time, multi-frame restart analysis and easy-convergence.

cond-mat.supr-con↗
arXiv2020

Analysis of Gender Inequality In Face Recognition Accuracy

We present a comprehensive analysis of how and why face recognition accuracy differs between men and women. We show that accuracy is lower for women due to the combination of (1) the impostor distribution for women having a skew toward higher similarity scores, and (2) the genuine distribution for women having a skew toward lower similarity scores. We show that this phenomenon of the impostor and genuine distributions for women shifting closer towards each other is general across datasets of African-American, Caucasian, and Asian faces. We show that the distribution of facial expressions may differ between male/female, but that the accuracy difference persists for image subsets rated confidently as neutral expression. The accuracy difference also persists for image subsets rated as close to zero pitch angle. Even when removing images with forehead partially occluded by hair/hat, the same impostor/genuine accuracy difference persists. We show that the female genuine distribution improves when only female images without facial cosmetics are used, but that the female impostor distribution also degrades at the same time. Lastly, we show that the accuracy difference persists even if a state-of-the-art deep learning method is trained from scratch using training data explicitly balanced between male and female images and subjects.

cs.CV↗
arXiv2019

Boundary Lipschitz Regularity and the Hopf Lemma on Reifenberg Domains for Fully Nonlinear Elliptic Equations

In this paper, we prove the boundary Lipschitz regularity and the Hopf Lemma by a unified method on Reifenberg domains for fully nonlinear elliptic equations. Precisely, if the domain $Ω$ satisfies the exterior Reifenberg $C^{1,\mathrm{Dini}}$ condition at $x_0\in \partial Ω$ (see Definition 1.3), the solution is Lipschitz continuous at $x_0$; if $Ω$ satisfies the interior Reifenberg $C^{1,\mathrm{Dini}}$ condition at $x_0$ (see Definition 1.4), the Hopf lemma holds at $x_0$. Our paper extends the results under the usual $C^{1,\mathrm{Dini}}$ condition.

math.AP↗
arXiv2019

A Dynamic and Cooperative Tracking System for Crowdfunding

Crowdfunding is an emerging finance platform for creators to fund their efforts by soliciting relatively small contributions from a large number of individuals using the Internet. Due to the unique rules, a campaign succeeds in trading only when it collects adequate funds in a given time. To prevent creators and backers from wasting time and efforts on failing campaigns, dynamically estimating the success probability of a campaign is very important. However, existing crowdfunding systems neither have the mechanism of dynamic predictive tracking, nor provide the real-time campaign status for creators and backers on the platform. To address these issues, we develop a novel system, which contains a dynamic data-driven approach to tracking the success probability and status. We demonstrate the following scenarios using our system. First, users can utilize our system to analyze the emotion of incremental reviews so as to understand backers' perspectives of the campaign in time. Meanwhile, our system visualizes the statistic number of positive and negative reviews. On this basis, our system can dynamically track the success probability of each campaign.

cs.SI↗
arXiv2019

Machine learning based data retrieval for inverse scattering problems with incomplete data

We are concerned with the inverse scattering problems associated with incomplete measurement data. It is a challenging topic of increasing importance in many practical applications. Based on a prototypical working model, we propose a machine learning based inverse scattering scheme, which integrates a CNN (convolution neural network) for the data retrieval. The proposed method can effectively cope with the reconstruction under limited-aperture and/or phaseless far-field data. Numerical experiments verify the promising features of our new scheme.

math.AP↗
arXiv2019

AIM 2019 Challenge on Constrained Super-Resolution: Methods and Results

This paper reviews the AIM 2019 challenge on constrained example-based single image super-resolution with focus on proposed solutions and results. The challenge had 3 tracks. Taking the three main aspects (i.e., number of parameters, inference/running time, fidelity (PSNR)) of MSRResNet as the baseline, Track 1 aims to reduce the amount of parameters while being constrained to maintain or improve the running time and the PSNR result, Tracks 2 and 3 aim to optimize running time and PSNR result with constrain of the other two aspects, respectively. Each track had an average of 64 registered participants, and 12 teams submitted the final results. They gauge the state-of-the-art in single image super-resolution.

eess.IV↗
arXiv2019

Unsupervised Context Rewriting for Open Domain Conversation

Context modeling has a pivotal role in open domain conversation. Existing works either use heuristic methods or jointly learn context modeling and response generation with an encoder-decoder framework. This paper proposes an explicit context rewriting method, which rewrites the last utterance by considering context history. We leverage pseudo-parallel data and elaborate a context rewriting network, which is built upon the CopyNet with the reinforcement learning method. The rewritten utterance is beneficial to candidate retrieval, explainable context modeling, as well as enabling to employ a single-turn framework to the multi-turn scenario. The empirical results show that our model outperforms baselines in terms of the rewriting quality, the multi-turn response generation, and the end-to-end retrieval-based chatbots.

cs.CL↗
arXiv2019

Leveraging Vision Reconstruction Pipelines for Satellite Imagery

Reconstructing 3D geometry from satellite imagery is an important topic of research. However, disparities exist between how this 3D reconstruction problem is handled in the remote sensing context and how multi-view reconstruction pipelines have been developed in the computer vision community. In this paper, we explore whether state-of-the-art reconstruction pipelines from the vision community can be applied to the satellite imagery. Along the way, we address several challenges adapting vision-based structure from motion and multi-view stereo methods. We show that vision pipelines can offer competitive speed and accuracy in the satellite context.

cs.CV↗
arXiv2019

Exploring Overall Contextual Information for Image Captioning in Human-Like Cognitive Style

Image captioning is a research hotspot where encoder-decoder models combining convolutional neural network (CNN) and long short-term memory (LSTM) achieve promising results. Despite significant progress, these models generate sentences differently from human cognitive styles. Existing models often generate a complete sentence from the first word to the end, without considering the influence of the following words on the whole sentence generation. In this paper, we explore the utilization of a human-like cognitive style, i.e., building overall cognition for the image to be described and the sentence to be constructed, for enhancing computer image understanding. This paper first proposes a Mutual-aid network structure with Bidirectional LSTMs (MaBi-LSTMs) for acquiring overall contextual information. In the training process, the forward and backward LSTMs encode the succeeding and preceding words into their respective hidden states by simultaneously constructing the whole sentence in a complementary manner. In the captioning process, the LSTM implicitly utilizes the subsequent semantic information contained in its hidden states. In fact, MaBi-LSTMs can generate two sentences in forward and backward directions. To bridge the gap between cross-domain models and generate a sentence with higher quality, we further develop a cross-modal attention mechanism to retouch the two sentences by fusing their salient parts as well as the salient areas of the image. Experimental results on the Microsoft COCO dataset show that the proposed model improves the performance of encoder-decoder models and achieves state-of-the-art results.

cs.CV↗
arXiv2019

Generalizable Meta-Heuristic based on Temporal Estimation of Rewards for Large Scale Blackbox Optimization

The generalization abilities of heuristic optimizers may deteriorate with the increment of the search space dimensionality. To achieve generalized performance across Large Scale Blackbox Optimization (LSBO) tasks, it ispossible to ensemble several heuristics and devise a meta-heuristic to control their initiation. This paper first proposes a methodology of transforming LSBO problems into online decision processes to maximize efficiency of resource utilization. Then, using the perspective of multi-armed bandits with non-stationary reward distributions, we propose a meta-heuristic based on Temporal Estimation of Rewards (TER) to address such decision process. TER uses a window for temporal credit assignment and Boltzmann exploration to balance the exploration-exploitation tradeoff. The prior-free TER generalizes across LSBO tasks with flexibility for different types of limited computational resources (e.g. time, money, etc.) and is easy to be adapted to new tasks for its simplicity and easy interface for heuristic articulation. Tests on the benchmarks validate the problem formulation and suggest significant effectiveness: when TER is articulated with three heuristics, competitive performance is reported across different sets of benchmark problems with search dimensions up to 10000.

cs.NE↗
arXiv2019

Bistability in the wake of a wavy cylinder

We investigate the wake dynamics of an optimally designed wavy cylinder that completely suppresses the Kármán vortex shedding. Such a wavy cylinder is forced to oscillate with a sinusoidal motion in the crossflow direction. Examination of the lift force spectrum reveals that a critical forcing frequency exists, below which the flow control effectiveness of the wavy cylinder is retained, and beyond which the inherent vortex shedding resurrects. Moreover, the resurrected unsteady vortex shedding can persist even without sustained forcing, leading to the loss of control efficacy. This suggests that in addition to the steady state developed from uniform initial condition, an oscillatory state exists in the wake of wavy cylinder if the initial state is sufficiently perturbed. The discovery of the bistable states calls for examination of the flow control effectiveness of the wavy cylinder in more complicated inflow conditions.

physics.flu-dyn↗
arXiv2019

On the formation of three-dimensional flows over finite-aspect-ratio wings under tip effects

We perform DNS of flow over finite-aspect-ratio NACA 0015 wings to characterize the tip effects on the wake dynamics. This study focuses on the development of three-dimensional separated flow over the wing, and discuss flow structures formed on the wing surface as well as in the far field wake. Vorticity is introduced into the flow in a predominantly two-dimensional manner. The vortex sheet from the wing tip rolls up around the free end to form the tip vortex. At its inception, the tip vortex is weak and its effect is spatially confined. As the flow around the tip separates, the tip effects extend farther in the spanwise direction, generating noticeable three dimensionality in the wake. For low-aspect-ratio wings, the wake remains stable due to the strong downwash over the entire span. Increasing the aspect ratio allows unsteady vortical flow to emerge away from the tip at sufficiently high angles of attack. These unsteady vortices shed and form closed vortical loops. For higher-aspect-ratio wings, the tip effects retard the near-tip shedding process, which desynchronizes from the two-dimensional shedding over the mid-span region, yielding vortex dislocation. At high angles of attack, the tip vortex exhibits noticeable undulations due to the strong interaction with the unsteady shedding vortices. The spanwise distribution of force coefficients is related to the three-dimensional wake dynamics and the tip effects. Vortical elements in the wake that are responsible for the generation of lift and drag forces are identified through the force element analysis. We note that at high angles of attack, a stationary vortical structure forms at the leading edge near the tip, giving rise to locally high lift and drag forces. The analysis performed here reveals how the vortical flow around the tip influences the separation physics, the global wake dynamics, and the spanwise force distributions.

physics.flu-dyn↗
arXiv2019

Variable screening based on Gaussian Centered L-moments

An important challenge in big data is identification of important variables. In this paper, we propose methods of discovering variables with non-standard univariate marginal distributions. The conventional moments-based summary statistics can be well-adopted for that purpose, but their sensitivity to outliers can lead to selection based on a few outliers rather than distributional shape such as bimodality. To address this type of non-robustness, we consider the L-moments. Using these in practice, however, has a limitation because they do not take zero values at the Gaussian distributions to which the shape of a marginal distribution is most naturally compared. As a remedy, we propose Gaussian Centered L-moments which share advantages of the L-moments but have zeros at the Gaussian distributions. The strength of Gaussian Centered L-moments over other conventional moments is shown in theoretical and practical aspects such as their performances in screening important genes in cancer genetics data.

stat.ME↗