arXiv subjects

Jia Guo

Publications and source records attributed to Jia Guo.

At least 127 records · Page 7Linked to original sources

arXiv2020

LAMP: Label Augmented Multimodal Pretraining

Multi-modal representation learning by pretraining has become an increasing interest due to its easy-to-use and potential benefit for various Visual-and-Language~(V-L) tasks. However its requirement of large volume and high-quality vision-language pairs highly hinders its values in practice. In this paper, we proposed a novel label-augmented V-L pretraining model, named LAMP, to address this problem. Specifically, we leveraged auto-generated labels of visual objects to enrich vision-language pairs with fine-grained alignment and correspondingly designed a novel pretraining task. Besides, we also found such label augmentation in second-stage pretraining would further universally benefit various downstream tasks. To evaluate LAMP, we compared it with some state-of-the-art models on four downstream tasks. The quantitative results and analysis have well proven the value of labels in V-L pretraining and the effectiveness of LAMP.

cs.MM↗

arXiv2020

Learning Power Control for Cellular Systems with Heterogeneous Graph Neural Network

Optimizing power control in multi-cell cellular networks with deep learning enables such a non-convex problem to be implemented in real-time. When channels are time-varying, the deep neural networks (DNNs) need to be re-trained frequently, which calls for low training complexity. To reduce the number of training samples and the size of DNN required to achieve good performance, a promising approach is to embed the DNNs with priori knowledge. Since cellular networks can be modelled as a graph, it is natural to employ graph neural networks (GNNs) for learning, which exhibit permutation invariance (PI) and equivalence (PE) properties. Unlike the homogeneous GNNs that have been used for wireless problems, whose outputs are invariant or equivalent to arbitrary permutations of vertexes, heterogeneous GNNs (HetGNNs), which are more appropriate to model cellular networks, are only invariant or equivalent to some permutations. If the PI or PE properties of the HetGNN do not match the property of the task to be learned, the performance degrades dramatically. In this paper, we show that the power control policy has a combination of different PI and PE properties, and existing HetGNN does not satisfy these properties. We then design a parameter sharing scheme for HetGNN such that the learned relationship satisfies the desired properties. Simulation results show that the sample complexity and the size of designed GNN for learning the optimal power control policy in multi-user multi-cell networks are much lower than the existing DNNs, when achieving the same sum rate loss from the numerically obtained solutions.

cs.LG↗

arXiv2020

Sufficient Conditions for Parameter Convergence over Embedded Manifolds using Kernel Techniques

The persistence of excitation (PE) condition is sufficient to ensure parameter convergence in adaptive estimation problems. Recent results on adaptive estimation in reproducing kernel Hilbert spaces (RKHS) introduce PE conditions for RKHS. This paper presents sufficient conditions for PE for the particular class of uniformly embedded reproducing kernel Hilbert spaces (RKHS) defined over smooth Riemannian manifolds. This paper also studies the implications of the sufficient condition in the case when the RKHS is finite or infinite-dimensional. When the RKHS is finite-dimensional, the sufficient condition implies parameter convergence as in the conventional analysis. On the other hand, when the RKHS is infinite-dimensional, the same condition implies that the function estimate error is ultimately bounded by a constant that depends on the approximation error in the infinite-dimensional RKHS. We illustrate the effectiveness of the sufficient condition in a practical example.

eess.SY↗

arXiv2020

Kernel Center Adaptation in the Reproducing Kernel Hilbert Space Embedding Method

The performance of adaptive estimators that employ embedding in reproducing kernel Hilbert spaces (RKHS) depends on the choice of the location of basis kernel centers. Parameter convergence and error approximation rates depend on where and how the kernel centers are distributed in the state-space. In this paper, we develop the theory that relates parameter convergence and approximation rates to the position of kernel centers. We develop criteria for choosing kernel centers in a specific class of systems - ones in which the state trajectory regularly visits the neighborhood of the positive limit set. Two algorithms, based on centroidal Voronoi tessellations and Kohonen self-organizing maps, are derived to choose kernel centers in the RKHS embedding method. Finally, we implement these methods on two practical examples and test their effectiveness.

eess.SY↗

arXiv2020

Darwin's Neural Network: AI-based Strategies for Rapid and Scalable Cell and Coronavirus Screening

Recent advances in the interdisciplinary scientific field of machine perception, computer vision, and biomedical engineering underpin a collection of machine learning algorithms with a remarkable ability to decipher the contents of microscope and nanoscope images. Machine learning algorithms are transforming the interpretation and analysis of microscope and nanoscope imaging data through use in conjunction with biological imaging modalities. These advances are enabling researchers to carry out real-time experiments that were previously thought to be computationally impossible. Here we adapt the theory of survival of the fittest in the field of computer vision and machine perception to introduce a new framework of multi-class instance segmentation deep learning, Darwin's Neural Network (DNN), to carry out morphometric analysis and classification of COVID19 and MERS-CoV collected in vivo and of multiple mammalian cell types in vitro.

eess.IV↗

arXiv2020

Approximations of the Reproducing Kernel Hilbert Space (RKHS) Embedding Method over Manifolds

The reproducing kernel Hilbert space (RKHS) embedding method is a recently introduced estimation approach that seeks to identify the unknown or uncertain function in the governing equations of a nonlinear set of ordinary differential equations (ODEs). While the original state estimate evolves in Euclidean space, the function estimate is constructed in an infinite-dimensional RKHS that must be approximated in practice. When a finite-dimensional approximation is constructed using a basis defined in terms of shifted kernel functions centered at the observations along a trajectory, the RKHS embedding method can be understood as a data-driven approach. This paper derives sufficient conditions that ensure that approximations of the unknown function converge in a Sobolev norm over a submanifold that supports the dynamics. Moreover, the rate of convergence for the finite-dimensional approximations is derived in terms of the fill distance of the samples in the embedded manifold. Numerical simulation of an example problem is carried out to illustrate the qualitative nature of convergence results derived in the paper.

math.OC↗

arXiv2020

Continuity of Utility Maximization under Weak Convergence

In this paper we find tight sufficient conditions for the continuity of the value of the utility maximization problem from terminal wealth with respect to the convergence in distribution of the underlying processes. We also establish a weak convergence result for the terminal wealths of the optimal portfolios. Finally, we apply our results to the computation of the minimal expected shortfall (shortfall risk) in the Heston model by building an appropriate lattice approximation.

q-fin.MF↗

arXiv2020

Intrinsic and Extrinsic Approximation of Koopman Operators over Manifolds

This paper derives rates of convergence of certain approximations of the Koopman operators that are associated with discrete, deterministic, continuous semiflows on a complete metric space $(X,d_X)$. Approximations are constructed in terms of reproducing kernel bases that are centered at samples taken along the system trajectory. It is proven that when the samples are dense in a certain type of smooth manifold $M\subseteq X$, the derived rates of convergence depend on the fill distance of samples along the trajectory in that manifold. Error bounds for projection-based and data-dependent approximations of the Koopman operator are derived in the paper. A discussion of how these bounds are realized in intrinsic and extrinsic approximation methods is given. Finally, a numerical example that illustrates qualitatively the convergence guarantees derived in the paper is given.

math.DS↗

arXiv2020

RKHS Embedding for Estimating Nonlinear Piezoelectric Systems

Nonlinearities in piezoelectric systems can arise from internal factors such as nonlinear constitutive laws or external factors like realizations of boundary conditions. It can be difficult or even impossible to derive detailed models from the first principles of all the sources of nonlinearity in a system. As a specific example, in traditional modeling techniques that use electric enthalpy density with higher-order terms, it can be problematic to choose which polynomial nonlinearities are essential. This paper introduces adaptive estimator techniques to estimate the nonlinearities that can arise in certain piezoelectric systems. Here an underlying assumption is that the nonlinearities can be modeled as functions in a reproducing kernel Hilbert space (RKHS). Unlike traditional modeling approaches, the approach discussed in this paper allows the development of models without knowledge of the precise form or structure of the nonlinearity. This approach can be viewed as a data-driven method to approximate the unknown nonlinear system. This paper introduces the theory behind the adaptive estimator and studies the effectiveness of this approach numerically for a class of nonlinear piezoelectric composite beams.

math.OC↗

arXiv2020

Constructing Deep Neural Networks with a Priori Knowledge of Wireless Tasks

Deep neural networks (DNNs) have been employed for designing wireless systems in many aspects, say transceiver design, resource optimization, and information prediction. Existing works either use the fully-connected DNN or the DNNs with particular architectures developed in other domains. While generating labels for supervised learning and gathering training samples are time-consuming or cost-prohibitive, how to develop DNNs with wireless priors for reducing training complexity remains open. In this paper, we show that two kinds of permutation invariant properties widely existed in wireless tasks can be harnessed to reduce the number of model parameters and hence the sample and computational complexity for training. We find special architecture of DNNs whose input-output relationships satisfy the properties, called permutation invariant DNN (PINN), and augment the data with the properties. By learning the impact of the scale of a wireless system, the size of the constructed PINNs can flexibly adapt to the input data dimension. We take predictive resource allocation and interference coordination as examples to show how the PINNs can be employed for learning the optimal policy with unsupervised and supervised learning. Simulations results demonstrate a dramatic gain of the proposed PINNs in terms of reducing training complexity.

cs.LG↗

arXiv2020

Segmentation with Residual Attention U-Net and an Edge-Enhancement Approach Preserves Cell Shape Features

The ability to extrapolate gene expression dynamics in living single cells requires robust cell segmentation, and one of the challenges is the amorphous or irregularly shaped cell boundaries. To address this issue, we modified the U-Net architecture to segment cells in fluorescence widefield microscopy images and quantitatively evaluated its performance. We also proposed a novel loss function approach that emphasizes the segmentation accuracy on cell boundaries and encourages shape feature preservation. With a 97% sensitivity, 93% specificity, 91% Jaccard similarity, and 95% Dice coefficient, our proposed method called Residual Attention U-Net with edge-enhancement surpassed the state-of-the-art U-Net in segmentation performance as evaluated by the traditional metrics. More remarkably, the same proposed candidate also performed the best in terms of the preservation of valuable shape features, namely area, eccentricity, major axis length, solidity and orientation. These improvements on shape feature preservation can serve as useful assets for downstream cell tracking and quantification of changes in cell statistics or features over time.

q-bio.QM↗

arXiv2020

Substituting Gadolinium in Brain MRI Using DeepContrast

Cerebral blood volume (CBV) is a hemodynamic correlate of oxygen metabolism and reflects brain activity and function. High-resolution CBV maps can be generated using the steady-state gadolinium-enhanced MRI technique. Such a technique requires an intravenous injection of exogenous gadolinium based contrast agent (GBCA) and recent studies suggest that the GBCA can accumulate in the brain after frequent use. We hypothesize that endogenous sources of contrast might exist within the most conventional and commonly acquired structural MRI, potentially obviating the need for exogenous contrast. Here, we test this hypothesis by developing and optimizing a deep learning algorithm, which we call DeepContrast, in mice. We find that DeepContrast performs equally well as exogenous GBCA in mapping CBV of the normal brain tissue and enhancing glioblastoma. Together, these studies validate our hypothesis that a deep learning approach can potentially replace the need for GBCAs in brain MRI.

q-bio.QM↗

arXiv2020

Edge-fault-tolerant strong Menger edge connectivity of bubble-sort star graphs

The connectivity and edge connectivity of interconnection network determine the fault tolerance of the network. An interconnection network is usually viewed as a connected graph, where vertex corresponds processor and edge corresponds link between two distinct processors. Given a connected graph $G$ with vertex set $V(G)$ and edge set $E(G)$, if for any two distinct vertices $u,v\in V(G)$, there exist $\min\{d_G(u),d_G(v)\}$ edge-disjoint paths between $u$ and $v$, then $G$ is strongly Menger edge connected. Let $m$ be an integer with $m\geq1$. If $G-F_e$ remains strongly Menger edge connected for any $F_e\subseteq E(G)$ with $|F_e|\leq m$, then $G$ is $m$-edge-fault-tolerant strongly Menger edge connected. If $G-F_e$ is strongly Menger edge connected for any $F_e\subseteq E(G)$ with $|F_e|\leq m$ and $δ(G-F_e)\geq2$, then $G$ is $m$-conditional edge-fault-tolerant strongly Menger edge connected. In this paper, we consider the $n$-dimensional bubble-sort star graph $BS_n$. We show that $BS_n$ is $(2n-5)$-edge-fault-tolerant strongly Menger edge connected for $n\geq3$ and $(6n-17)$-conditional edge-fault-tolerant strongly Menger edge connected for $n\geq4$. Moreover, we give some examples to show that our results are optimal.

math.CO↗

arXiv2019

Automatic Calculation of Resolution in Lateral Cephalogram Based on Scale Mark Detection

Cephalometric analysis is an important tool for orthodontic diagnosis. At present, most cephalometric analysis is performed with the help of image processing techniques. Hence, the resolution between millimeter and pixel is needed with high accuracy. In cephalometric analysis, a scale is placed in front of patient's head when taking the radiograph. This study aims to develop an algorithm to recognize the scale in cephalogram, locate the scale mark and calculate the pixel-millimeter-ratio. First, a ROI is detected and cropped based on regression tree voting. Second, an algorithm is employed in ROI to detect the corner points of the scale and rotate the scale to perfectly vertical direction. Finally, a pixel tracing algorithm is employed to locate the first and the last scale mark in order to calculate the pixel length of the calibration. A novel method is proposed to adaptively assign the size and orientation of the image patches described by the SIFT vectors to ensure invariance of scale and rotation of images. The algorithm is robust to interference including tags, stickers and stain. A dataset consisting 163 cephalograms is tested and the algorithm performs a 100% Success Detection Rate within the precision range of 1.0mm.

eess.IV↗

arXiv2019

Structure of Deep Neural Networks with a Priori Information in Wireless Tasks

Deep neural networks (DNNs) have been employed for designing wireless networks in many aspects, such as transceiver optimization, resource allocation, and information prediction. Existing works either use fully-connected DNN or the DNNs with specific structures that are designed in other domains. In this paper, we show that a priori information widely existed in wireless tasks is permutation invariant. For these tasks, we propose a DNN with special structure, where the weight matrices between layers of the DNN only consist of two smaller sub-matrices. By such way of parameter sharing, the number of model parameters reduces, giving rise to low sample and computational complexity for training a DNN. We take predictive resource allocation as an example to show how the designed DNN can be applied for learning the optimal policy with unsupervised learning. Simulations results validate our analysis and show dramatic gain of the proposed structure in terms of reducing training complexity.

cs.LG↗

arXiv2019

Persistence of Excitation in Reproducing Kernel Hilbert Spaces, Positive Limit Sets, and Smooth Manifolds

This paper studies the relationship between the positive limit sets of continuous semiflows and the newly introduced definition of persistently excited (PE) sets and associated subspaces of reproducing kernel Hilbert (RKH) spaces. It is shown that if the RKH space contains a rich collection of cut-off functions, persistently excited sets are contained as subsets of the positive limit set of the semiflow. The paper demonstrates how the new PE condition can be used to guarantee convergence of function estimates in the RKH space embedding method for adaptive estimation. In particular, the paper is applied to uncertain ODE systems with positive limit sets given by certain types of smooth manifolds, and it establishes convergence of adaptive function estimates over the manifolds.

math.OC↗

arXiv2019

Edge-bipancyclicity of bubble-sort star graphs

The interconnection network considered in this paper is the bubble-sort star graph. The $n$-dimensional bubble-sort star graph $BS_n$ is a bipartite and $(2n-3)$-regular graph of order $n!$. A bipartite graph $G$ is edge-bipancyclic if each edge of $G$ lies on a cycle of all even length $l$ with $4\leq l\leq |V(G)|$. In this paper, we show that the $n$-dimensional bubble-sort star graph $BS_n$ is edge-bipancyclic for $n\ge 3$ and for each even length $l$ with $4\leq l\leq n!$, every edge of $BS_n$ lies on at least four different cycles of length $l$.

math.CO↗

arXiv2019

RetinaFace: Single-stage Dense Face Localisation in the Wild

Though tremendous strides have been made in uncontrolled face detection, accurate and efficient face localisation in the wild remains an open challenge. This paper presents a robust single-stage face detector, named RetinaFace, which performs pixel-wise face localisation on various scales of faces by taking advantages of joint extra-supervised and self-supervised multi-task learning. Specifically, We make contributions in the following five aspects: (1) We manually annotate five facial landmarks on the WIDER FACE dataset and observe significant improvement in hard face detection with the assistance of this extra supervision signal. (2) We further add a self-supervised mesh decoder branch for predicting a pixel-wise 3D shape face information in parallel with the existing supervised branches. (3) On the WIDER FACE hard test set, RetinaFace outperforms the state of the art average precision (AP) by 1.1% (achieving AP equal to 91.4%). (4) On the IJB-C test set, RetinaFace enables state of the art methods (ArcFace) to improve their results in face verification (TAR=89.59% for FAR=1e-6). (5) By employing light-weight backbone networks, RetinaFace can run real-time on a single CPU core for a VGA-resolution image. Extra annotations and code have been made available at: https://github.com/deepinsight/insightface/tree/master/RetinaFace.

cs.CV↗