arXiv subjects

Hui Tang

Publications and source records attributed to Hui Tang.

At least 73 records · Page 4Linked to original sources

arXiv2020

Applying deep reinforcement learning to active flow control in turbulent conditions

Machine learning has recently become a promising technique in fluid mechanics, especially for active flow control (AFC) applications. A recent work [J. Fluid Mech. (2019), vol. 865, pp. 281-302] has demonstrated the feasibility and effectiveness of deep reinforcement learning (DRL) in performing AFC over a circular cylinder at $Re = 100$, i.e., in the laminar flow regime. As a follow-up study, we investigate the same AFC problem at an intermediate Reynolds number, i.e., $Re = 1000$, where the turbulence in the flow poses great challenges to the control. The results show that the DRL agent can still find effective control strategies, but requires much more episodes in the learning. A remarkable drag reduction of around $30\%$ is achieved, which is accompanied by elongation of the recirculation bubble and reduction of turbulent fluctuations in the cylinder wake. To our best knowledge, this study is the first successful application of DRL to AFC in weak turbulent conditions. It therefore sets a new milestone in progressing towards AFC in strong turbulent flows.

physics.flu-dyn↗

arXiv2020

Bipartite Distance for Shape-Aware Landmark Detection in Spinal X-Ray Images

Scoliosis is a congenital disease that causes lateral curvature in the spine. Its assessment relies on the identification and localization of vertebrae in spinal X-ray images, conventionally via tedious and time-consuming manual radiographic procedures that are prone to subjectivity and observational variability. Reliability can be improved through the automatic detection and localization of spinal landmarks. To guide a CNN in the learning of spinal shape while detecting landmarks in X-ray images, we propose a novel loss based on a bipartite distance (BPD) measure, and show that it consistently improves landmark detection performance.

eess.IV↗

arXiv2020

Partly Supervised Multitask Learning

Semi-supervised learning has recently been attracting attention as an alternative to fully supervised models that require large pools of labeled data. Moreover, optimizing a model for multiple tasks can provide better generalizability than single-task learning. Leveraging self-supervision and adversarial training, we propose a novel general purpose semi-supervised, multiple-task model---namely, self-supervised, semi-supervised, multitask learning (S$^4$MTL)---for accomplishing two important tasks in medical imaging, segmentation and diagnostic classification. Experimental results on chest and spine X-ray datasets suggest that our S$^4$MTL model significantly outperforms semi-supervised single task, semi/fully-supervised multitask, and fully-supervised single task models, even with a 50\% reduction of class and segmentation labels. We hypothesize that our proposed model can be effective in tackling limited annotation problems for joint training, not only in medical imaging domains, but also for general-purpose vision tasks.

cs.CV↗

arXiv2020

Analysis of Scoliosis From Spinal X-Ray Images

Scoliosis is a congenital disease in which the spine is deformed from its normal shape. Measurement of scoliosis requires labeling and identification of vertebrae in the spine. Spine radiographs are the most cost-effective and accessible modality for imaging the spine. Reliable and accurate vertebrae segmentation in spine radiographs is crucial in image-guided spinal assessment, disease diagnosis, and treatment planning. Conventional assessments rely on tedious and time-consuming manual measurement, which is subject to inter-observer variability. A fully automatic method that can accurately identify and segment the associated vertebrae is unavailable in the literature. Leveraging a carefully-adjusted U-Net model with progressive side outputs, we propose an end-to-end segmentation model that provides a fully automatic and reliable segmentation of the vertebrae associated with scoliosis measurement. Our experimental results from a set of anterior-posterior spine X-Ray images indicate that our model, which achieves an average Dice score of 0.993, promises to be an effective tool in the identification and labeling of spinal vertebrae, eventually helping doctors in the reliable estimation of scoliosis. Moreover, estimation of Cobb angles from the segmented vertebrae further demonstrates the effectiveness of our model.

eess.IV↗

arXiv2020

Unsupervised Domain Adaptation via Structurally Regularized Deep Clustering

Unsupervised domain adaptation (UDA) is to make predictions for unlabeled data on a target domain, given labeled data on a source domain whose distribution shifts from the target one. Mainstream UDA methods learn aligned features between the two domains, such that a classifier trained on the source features can be readily applied to the target ones. However, such a transferring strategy has a potential risk of damaging the intrinsic discrimination of target data. To alleviate this risk, we are motivated by the assumption of structural domain similarity, and propose to directly uncover the intrinsic target discrimination via discriminative clustering of target data. We constrain the clustering solutions using structural source regularization that hinges on our assumed structural domain similarity. Technically, we use a flexible framework of deep network based discriminative clustering that minimizes the KL divergence between predictive label distribution of the network and an introduced auxiliary one; replacing the auxiliary distribution with that formed by ground-truth labels of source data implements the structural source regularization via a simple strategy of joint network training. We term our proposed method as Structurally Regularized Deep Clustering (SRDC), where we also enhance target discrimination with clustering of intermediate network features, and enhance structural regularization with soft selection of less divergent source examples. Careful ablation studies show the efficacy of our proposed SRDC. Notably, with no explicit domain alignment, SRDC outperforms all existing methods on three UDA benchmarks.

cs.CV↗

arXiv2020

Deep Reinforcement Learning in Fluid Mechanics: a promising method for both Active Flow Control and Shape Optimization

In recent years, Artificial Neural Networks (ANNs) and Deep Learning have become increasingly popular across a wide range of scientific and technical fields, including Fluid Mechanics. While it will take time to fully grasp the potentialities as well as the limitations of these methods, evidence is starting to accumulate that point to their potential in helping solve problems for which no theoretically optimal solution method is known. This is particularly true in Fluid Mechanics, where problems involving optimal control and optimal design are involved. Indeed, such problems are famously difficult to solve effectively with traditional methods due to the combination of non linearity, non convexity, and high dimensionality they involve. By contrast, Deep Reinforcement Learning (DRL), a method of optimization based on teaching empirical strategies to an ANN through trial and error, is well adapted to solving such problems. In this short review, we offer an insight into the current state of the art of the use of DRL within fluid mechanics, focusing on control and optimal design problems.

physics.flu-dyn↗

arXiv2019

Discriminative Adversarial Domain Adaptation

Given labeled instances on a source domain and unlabeled ones on a target domain, unsupervised domain adaptation aims to learn a task classifier that can well classify target instances. Recent advances rely on domain-adversarial training of deep networks to learn domain-invariant features. However, due to an issue of mode collapse induced by the separate design of task and domain classifiers, these methods are limited in aligning the joint distributions of feature and category across domains. To overcome it, we propose a novel adversarial learning method termed Discriminative Adversarial Domain Adaptation (DADA). Based on an integrated category and domain classifier, DADA has a novel adversarial objective that encourages a mutually inhibitory relation between category and domain predictions for any input instance. We show that under practical conditions, it defines a minimax game that can promote the joint distribution alignment. Except for the traditional closed set domain adaptation, we also extend DADA for extremely challenging problem settings of partial and open set domain adaptation. Experiments show the efficacy of our proposed methods and we achieve the new state of the art for all the three settings on benchmark datasets.

cs.CV↗

arXiv2019

Shape-Aware Organ Segmentation by Predicting Signed Distance Maps

In this work, we propose to resolve the issue existing in current deep learning based organ segmentation systems that they often produce results that do not capture the overall shape of the target organ and often lack smoothness. Since there is a rigorous mapping between the Signed Distance Map (SDM) calculated from object boundary contours and the binary segmentation map, we exploit the feasibility of learning the SDM directly from medical scans. By converting the segmentation task into predicting an SDM, we show that our proposed method retains superior segmentation performance and has better smoothness and continuity in shape. To leverage the complementary information in traditional segmentation training, we introduce an approximated Heaviside function to train the model by predicting SDMs and segmentation maps simultaneously. We validate our proposed models by conducting extensive experiments on a hippocampus segmentation dataset and the public MICCAI 2015 Head and Neck Auto Segmentation Challenge dataset with multiple organs. While our carefully designed backbone 3D segmentation network improves the Dice coefficient by more than 5% compared to current state-of-the-arts, the proposed model with SDM learning produces smoother segmentation results with smaller Hausdorff distance and average surface distance, thus proving the effectiveness of our method.

cs.CV↗

arXiv2019

Object-Guided Instance Segmentation for Biological Images

Instance segmentation of biological images is essential for studying object behaviors and properties. The challenges, such as clustering, occlusion, and adhesion problems of the objects, make instance segmentation a non-trivial task. Current box-free instance segmentation methods typically rely on local pixel-level information. Due to a lack of global object view, these methods are prone to over- or under-segmentation. On the contrary, the box-based instance segmentation methods incorporate object detection into the segmentation, performing better in identifying the individual instances. In this paper, we propose a new box-based instance segmentation method. Mainly, we locate the object bounding boxes from their center points. The object features are subsequently reused in the segmentation branch as a guide to separate the clustered instances within an RoI patch. Along with the instance normalization, the model is able to recover the target object distribution and suppress the distribution of neighboring attached objects. Consequently, the proposed model performs excellently in segmenting the clustered objects while retaining the target object details. The proposed method achieves state-of-the-art performances on three biological datasets: cell nuclei, plant phenotyping dataset, and neural cells.

cs.CV↗

arXiv2019

Domain-Symmetric Networks for Adversarial Domain Adaptation

Unsupervised domain adaptation aims to learn a model of classifier for unlabeled samples on the target domain, given training data of labeled samples on the source domain. Impressive progress is made recently by learning invariant features via domain-adversarial training of deep networks. In spite of the recent progress, domain adaptation is still limited in achieving the invariance of feature distributions at a finer category level. To this end, we propose in this paper a new domain adaptation method called Domain-Symmetric Networks (SymNets). The proposed SymNet is based on a symmetric design of source and target task classifiers, based on which we also construct an additional classifier that shares with them its layer neurons. To train the SymNet, we propose a novel adversarial learning objective whose key design is based on a two-level domain confusion scheme, where the category-level confusion loss improves over the domain-level one by driving the learning of intermediate network features to be invariant at the corresponding categories of the two domains. Both domain discrimination and domain confusion are implemented based on the constructed additional classifier. Since target samples are unlabeled, we also propose a scheme of cross-domain training to help learn the target classifier. Careful ablation studies show the efficacy of our proposed method. In particular, based on commonly used base networks, our SymNets achieve the new state of the art on three benchmark domain adaptation datasets.

cs.CV↗

arXiv2018

3D Segmentation with Exponential Logarithmic Loss for Highly Unbalanced Object Sizes

With the introduction of fully convolutional neural networks, deep learning has raised the benchmark for medical image segmentation on both speed and accuracy, and different networks have been proposed for 2D and 3D segmentation with promising results. Nevertheless, most networks only handle relatively small numbers of labels (<10), and there are very limited works on handling highly unbalanced object sizes especially in 3D segmentation. In this paper, we propose a network architecture and the corresponding loss function which improve segmentation of very small structures. By combining skip connections and deep supervision with respect to the computational feasibility of 3D segmentation, we propose a fast converging and computationally efficient network architecture for accurate segmentation. Furthermore, inspired by the concept of focal loss, we propose an exponential logarithmic loss which balances the labels not only by their relative sizes but also by their segmentation difficulties. We achieve an average Dice coefficient of 82% on brain segmentation with 20 labels, with the ratio of the smallest to largest object sizes as 0.14%. Less than 100 epochs are required to reach such accuracy, and segmenting a 128x128x128 volume only takes around 0.4 s.

cs.CV↗

arXiv2018

Fine-Grained Visual Categorization using Meta-Learning Optimization with Sample Selection of Auxiliary Data

Fine-grained visual categorization (FGVC) is challenging due in part to the fact that it is often difficult to acquire an enough number of training samples. To employ large models for FGVC without suffering from overfitting, existing methods usually adopt a strategy of pre-training the models using a rich set of auxiliary data, followed by fine-tuning on the target FGVC task. However, the objective of pre-training does not take the target task into account, and consequently such obtained models are suboptimal for fine-tuning. To address this issue, we propose in this paper a new deep FGVC model termed MetaFGNet. Training of MetaFGNet is based on a novel regularized meta-learning objective, which aims to guide the learning of network parameters so that they are optimal for adapting to the target FGVC task. Based on MetaFGNet, we also propose a simple yet effective scheme for selecting more useful samples from the auxiliary data. Experiments on benchmark FGVC datasets show the efficacy of our proposed method.

cs.CV↗

arXiv2017

A Chunk Caching Location and Searching Scheme in Content Centric Networking

Content Centric Networking (CCN) is a new network infrastructure around content dissemination and retrieval, shift from host addresses to named data. Each CCN router has a cache to store the chunks passed by it. Therefore the caching strategy about chunk placement can greatly affect the whole CCN performance. This paper proposes an implicit coordinate chunk caching location and searching scheme (CLS) in CCN hierarchical infrastructure. In CLS, there is at most one copy of a chunk cached on the path between a server and a leaf router. This copy is pulled down one level towards the leaf router by a request or pushed up one level towards the server by the cache eviction. Thus, it is possible to store more diverse contents in the whole CCN and improve the network performance. Plus, in order to reduce the server workload and file download time, a caching trail of chunk is created to direct the following request where to find the chunk. Extensive test-bed experiments have been performed to evaluate the proposed scheme in terms of a wide range of performance metrics. The results show that the proposed scheme outperforms existing algorithms.

cs.NI↗

arXiv2016

GeoClaw-STRICHE: A coupled model for Sediment TRansport In Coastal Hazard Events

GeoClaw-STRICHE is designed for simulating the physical impacts of tsunami as it relates to erosion, transport and deposition. GeoClaw-STRICHE comprises of three components: (1) nonlinear shallow water equations; (2) advection-diffusion equation; (3) an equation for morphology updating. Multiple grain sizes and sediment layers are added into GeoClaw-STRICHE to simulate grain-size distribution and add the capability to develop grain-size trends from bottom to the top of a simulated deposit as well as along the inundation. Unlike previous models based on empirical equations or sediment concentration gradient, the standard Van Leer method is applied to calculate sediment flux. We tested and verified GeoClaw-STRICHE with flume experiment by \citet{johnson2016experimental} and data from the 2004 Indian Ocean tsunami in Kuala Meurisi as published in \citet{JGRF:JGRF786}. The comparison with experimental data shows GeoClaw-STRICHE's capability to simulate sediment thickness and grain-size distribution in experimental conditions. The agreement between experimental and model results is generally quite good, especially for sediment thickness, which builds confidence that sediment transport is correctly predicted by this model. The comparison with the data from the 2004 Indian Ocean tsunami reveals that the pattern of sediment thickness is well predicted and is of similar quality, if not better than the established computational models such as Delft3D.

physics.geo-ph↗

arXiv2016

TSUFLIND-EnKF inversion model applied to tsunami deposits for estimation of transient flow depth and speed with quantified uncertainties

Tsunami deposits are recordings of tsunami events that contain information about flow conditions. Deciphering quantitative information from tsunami deposits is especially important for analyzing paleo-tsunami events in which deposits comprise the only leftover physical evidence. The physical meaning of the deciphered quantities depends on the physical assumptions that are applied. The aim of our study is to estimate the characteristics of tsunamis and quantify the errors and uncertainties that inherent within them. To achieve this goal, we apply the TSUFLIND-EnKF inversion model to study the deposition of an idealized deposit created by a single tsunami wave and one real case from 2004 India ocean tsunami. TSUFLIND-EnKF model combines TSUFLIND for the deposition module with the Ensemble Kalman Filtering (EnKF) method. In our modeling, we assume that grain-size distribution and thickness from the idealized deposits at different depths can be used as an observational variable. Our tentative results indicate that sampling methods and sampling frequencies of tsunami deposits influence not only the magnitude of the inverted variables, but also their errors and uncertainties. An interesting result of our technique is that a larger number of samples from a given tsunami deposit does not automatically mean that the inversion results are more robust with smaller errors and decreased uncertainties.

physics.geo-ph↗

arXiv2016

Inversion of Tsunamis Characteristics from Sediment Deposits Based on Ensemble Kalman Filtering

Sediment deposits are the only leftover records from paleo tsunami events. Therefore, inverse modeling method based on the information contained in the deposit is an indispensable way of deciphering the quantitative characteristics of the tsunamis, e.g., the flow speed and the flow depth. While several models have been proposed to perform tsunami inversion, i.e., to infer the tsunami characteristics based on the sediment deposits, the existing methods lack mathematical rigorousness and are not able to account for uncertainties in the inferred quantities. In this work, we propose an inversion scheme based on Ensemble Kalman Filtering (EnKF) to infer tsunami characteristics from sediment deposits. In contrast to traditional data assimilation methods using EnKF, a novelty of the current work is that we augment the system state to include both the physical variables (sediment fluxes) that are observable and the unknown parameters (flow speed and flow depth) to be inferred. Based on the rigorous Bayesian inference theory, the inversion scheme provides quantified uncertainties on the inferred quantities, which clearly distinguishes the present method with existing schemes for tsunami inversion. Two test cases with synthetic observation data are used to verify the proposed inversion scheme. Numerical results show that the tsunami characteristics inferred from the sediment deposit information have a favorable agreement with the truths, which demonstrated the merits of the proposed tsunami inversion scheme.

physics.geo-ph↗

arXiv2015

A Model for TSUnami FLow INversion from Deposits (TSUFLIND)

Modern tsunami deposits are employed to estimate the overland flow characteristics of tsunamis. With the help of the overland-flow characteristics, the characteristics of the causative tsunami wave can be estimated. The understanding of tsunami deposits has tremendously improved over the last decades. There are three prominent inversion models: Moore advection model, Soulsby's model and TsuSedMod model. TSUFLIND incorporates all three models and adds new modules to better simulate tsunami deposit formation and calculate flow condition. TSUFLIND takes grain-size distribution, thickness, water depth and topography information as inputs. TSUFLIND computes sediment concentration, grain-size distribution of sediment source and initial flow condition to match the sediment thickness and grain size distribution from field observation. Furthermore, TSUFLIND estimates the flow speed, Froude number and representative wave amplitude. The model is tested by using field data collected at Ranganathapuram, India after the 2004 India Ocean tsunami. TSUFLIND reproduces the field measurement grain-size distribution with less than 5% error. Tsunami speed in this test case is about 4.7 m/s at 150 meters inland and decreases to 3.3 m/s 350 meters inland from the shoreline. The estimated wave amplitude of the largest wave for this test case is about 5 to 7 meters.

physics.geo-ph↗

arXiv2008

Classification of the Super-conducting order parameters under the point group symmetry for a multi-band system: application to LaOFeAs

All the possible super-conducting order parameters for the LaOFeAs system are classified by their transformation under the complete crystal symmetry. The general forms of the super-conducting gap functions for each class are discussed. We find that the gap functions in such a multi-band system belong to three types, full gap, nodal type and finite {}``Fermi arc'' type. Possible physical consequences caused by different types of gap functions are also discussed.

cond-mat.supr-con↗