arXiv subjects

Ping Zhong

Publications and source records attributed to Ping Zhong.

At least 55 records · Page 3Linked to original sources

arXiv2020

Few-shot Object Detection with Self-adaptive Attention Network for Remote Sensing Images

In remote sensing field, there are many applications of object detection in recent years, which demands a great number of labeled data. However, we may be faced with some cases where only limited data are available. In this paper, we proposed a few-shot object detector which is designed for detecting novel objects provided with only a few examples. Particularly, in order to fit the object detection settings, our proposed few-shot detector concentrates on the relations that lie in the level of objects instead of the full image with the assistance of Self-Adaptive Attention Network (SAAN). The SAAN can fully leverage the object-level relations through a relation GRU unit and simultaneously attach attention on object features in a self-adaptive way according to the object-level relations to avoid some situations where the additional attention is useless or even detrimental. Eventually, the detection results are produced from the features that are added with attention and thus are able to be detected simply. The experiments demonstrate the effectiveness of the proposed method in few-shot scenes.

cs.CV↗

arXiv2020

Few-shot Object Detection with Feature Attention Highlight Module in Remote Sensing Images

In recent years, there are many applications of object detection in remote sensing field, which demands a great number of labeled data. However, in many cases, data is extremely rare. In this paper, we proposed a few-shot object detector which is designed for detecting novel objects based on only a few examples. Through fully leveraging labeled base classes, our model that is composed of a feature-extractor, a feature attention highlight module as well as a two-stage detection backend can quickly adapt to novel classes. The pre-trained feature extractor whose parameters are shared produces general features. While the feature attention highlight module is designed to be light-weighted and simple in order to fit the few-shot cases. Although it is simple, the information provided by it in a serial way is helpful to make the general features to be specific for few-shot objects. Then the object-specific features are delivered to the two-stage detection backend for the detection results. The experiments demonstrate the effectiveness of the proposed method for few-shot cases.

cs.CV↗

arXiv2020

Statistical Loss and Analysis for Deep Learning in Hyperspectral Image Classification

Nowadays, deep learning methods, especially the convolutional neural networks (CNNs), have shown impressive performance on extracting abstract and high-level features from the hyperspectral image. However, general training process of CNNs mainly considers the pixel-wise information or the samples' correlation to formulate the penalization while ignores the statistical properties especially the spectral variability of each class in the hyperspectral image. These samples-based penalizations would lead to the uncertainty of the training process due to the imbalanced and limited number of training samples. To overcome this problem, this work characterizes each class from the hyperspectral image as a statistical distribution and further develops a novel statistical loss with the distributions, not directly with samples for deep learning. Based on the Fisher discrimination criterion, the loss penalizes the sample variance of each class distribution to decrease the intra-class variance of the training samples. Moreover, an additional diversity-promoting condition is added to enlarge the inter-class variance between different class distributions and this could better discriminate samples from different classes in hyperspectral image. Finally, the statistical estimation form of the statistical loss is developed with the training samples through multi-variant statistical analysis. Experiments over the real-world hyperspectral images show the effectiveness of the developed statistical loss for deep learning.

cs.CV↗

arXiv2019

Hyperspectral Image Classification With Context-Aware Dynamic Graph Convolutional Network

In hyperspectral image (HSI) classification, spatial context has demonstrated its significance in achieving promising performance. However, conventional spatial context-based methods simply assume that spatially neighboring pixels should correspond to the same land-cover class, so they often fail to correctly discover the contextual relations among pixels in complex situations, and thus leading to imperfect classification results on some irregular or inhomogeneous regions such as class boundaries. To address this deficiency, we develop a new HSI classification method based on the recently proposed Graph Convolutional Network (GCN), as it can flexibly encode the relations among arbitrarily structured non-Euclidean data. Different from traditional GCN, there are two novel strategies adopted by our method to further exploit the contextual relations for accurate HSI classification. First, since the receptive field of traditional GCN is often limited to fairly small neighborhood, we proposed to capture long range contextual relations in HSI by performing successive graph convolutions on a learned region-induced graph which is transformed from the original 2D image grids. Second, we refine the graph edge weight and the connective relationships among image regions by learning the improved adjacency matrix and the 'edge filter', so that the graph can be gradually refined to adapt to the representations generated by each graph convolutional layer. Such updated graph will in turn result in accurate region representations, and vice versa. The experiments carried out on three real-world benchmark datasets demonstrate that the proposed method yields significant improvement in the classification performance when compared with some state-of-the-art approaches.

cs.LG↗

arXiv2019

Diversity in Machine Learning

Machine learning methods have achieved good performance and been widely applied in various real-world applications. They can learn the model adaptively and be better fit for special requirements of different tasks. Generally, a good machine learning system is composed of plentiful training data, a good model training process, and an accurate inference. Many factors can affect the performance of the machine learning process, among which the diversity of the machine learning process is an important one. The diversity can help each procedure to guarantee a total good machine learning: diversity of the training data ensures that the training data can provide more discriminative information for the model, diversity of the learned model (diversity in parameters of each model or diversity among different base models) makes each parameter/model capture unique or complement information and the diversity in inference can provide multiple choices each of which corresponds to a specific plausible local optimal result. Even though the diversity plays an important role in machine learning process, there is no systematical analysis of the diversification in machine learning system. In this paper, we systematically summarize the methods to make data diversification, model diversification, and inference diversification in the machine learning process, respectively. In addition, the typical applications where the diversity technology improved the machine learning performance have been surveyed, including the remote sensing imaging tasks, machine translation, camera relocalization, image segmentation, object detection, topic modeling, and others. Finally, we discuss some challenges of the diversity technology in machine learning and point out some directions in future work.

cs.CV↗

arXiv2019

Multi-scale Dynamic Graph Convolutional Network for Hyperspectral Image Classification

Convolutional Neural Network (CNN) has demonstrated impressive ability to represent hyperspectral images and to achieve promising results in hyperspectral image classification. However, traditional CNN models can only operate convolution on regular square image regions with fixed size and weights, so they cannot universally adapt to the distinct local regions with various object distributions and geometric appearances. Therefore, their classification performances are still to be improved, especially in class boundaries. To alleviate this shortcoming, we consider employing the recently proposed Graph Convolutional Network (GCN) for hyperspectral image classification, as it can conduct the convolution on arbitrarily structured non-Euclidean data and is applicable to the irregular image regions represented by graph topological information. Different from the commonly used GCN models which work on a fixed graph, we enable the graph to be dynamically updated along with the graph convolution process, so that these two steps can be benefited from each other to gradually produce the discriminative embedded features as well as a refined graph. Moreover, to comprehensively deploy the multi-scale information inherited by hyperspectral images, we establish multiple input graphs with different neighborhood scales to extensively exploit the diversified spectral-spatial correlations at multiple scales. Therefore, our method is termed 'Multi-scale Dynamic Graph Convolutional Network' (MDGCN). The experimental results on three typical benchmark datasets firmly demonstrate the superiority of the proposed MDGCN to other state-of-the-art methods in both qualitative and quantitative aspects.

eess.IV↗

arXiv2019

A novel statistical metric learning for hyperspectral image classification

In this paper, a novel statistical metric learning is developed for spectral-spatial classification of the hyperspectral image. First, the standard variance of the samples of each class in each batch is used to decrease the intra-class variance within each class. Then, the distances between the means of different classes are used to penalize the inter-class variance of the training samples. Finally, the standard variance between the means of different classes is added as an additional diversity term to repulse different classes from each other. Experiments have conducted over two real-world hyperspectral image datasets and the experimental results have shown the effectiveness of the proposed statistical metric learning.

cs.CV↗

arXiv2019

An End-to-End Joint Unsupervised Learning of Deep Model and Pseudo-Classes for Remote Sensing Scene Representation

This work develops a novel end-to-end deep unsupervised learning method based on convolutional neural network (CNN) with pseudo-classes for remote sensing scene representation. First, we introduce center points as the centers of the pseudo classes and the training samples can be allocated with pseudo labels based on the center points. Therefore, the CNN model, which is used to extract features from the scenes, can be trained supervised with the pseudo labels. Moreover, a pseudo-center loss is developed to decrease the variance between the samples and the corresponding pseudo center point. The pseudo-center loss is important since it can update both the center points with the training samples and the CNN model with the center points in the training process simultaneously. Finally, joint learning of the pseudo-center loss and the pseudo softmax loss which is formulated with the samples and the pseudo labels is developed for unsupervised remote sensing scene representation to obtain discriminative representations from the scenes. Experiments are conducted over two commonly used remote sensing scene datasets to validate the effectiveness of the proposed method and the experimental results show the superiority of the proposed method when compared with other state-of-the-art methods.

cs.CV↗

arXiv2019

Quasifree stochastic cocycles and quantum random walks

The theory of quasifree quantum stochastic calculus for infinite-dimensional noise is developed within the framework of Hudson-Parthasarathy quantum stochastic calculus. The question of uniqueness for the covariance amplitude with respect to which a given unitary quantum stochastic cocycle is quasifree is addressed, and related to the minimality of the corresponding stochastic dilation. The theory is applied to the identification of a wide class of quantum random walks whose limit processes are driven by quasifree noises.

math-ph↗

arXiv2018

The PPT square conjecture holds generically for some classes of independent states

Let $|ψ\rangle\langle ψ|$ be a random pure state on $\mathbb{C}^{d^2}\otimes \mathbb{C}^s$, where $ψ$ is a random unit vector uniformly distributed on the sphere in $\mathbb{C}^{d^2}\otimes \mathbb{C}^s$. Let $ρ_1$ be random induced states $ρ_1=Tr_{\mathbb{C}^s}(|ψ\rangle\langle ψ|)$ whose distribution is $μ_{d^2,s}$; and let $ρ_2$ be random induced states following the same distribution $μ_{d^2,s}$ independent from $ρ_1$. Let $ρ$ be a random state induced by the entanglement swapping of $ρ_1$ and $ρ_2$. We show that the empirical spectrum of $ρ- {1\mkern -4mu{\rm l}}/d^2$ converges almost surely to the Marcenko-Pastur law with parameter $c^2$ as $d\rightarrow \infty$ and $s/d \rightarrow c$. As an application, we prove that the state $ρ$ is separable generically if $ρ_1, ρ_2$ are PPT entangled.

math-ph↗

arXiv2018

Massive MIMO relaying with linear precoding in correlated channels under limited feedback

In this paper we study on a massive MIMO relay system with linear precoding under the conditions of imperfect channel state information at the transmitter (CSIT) and per-user channel transmit correlation. In our system the source-relay channels are massive multiple-input multiple-output (MIMO) ones and the relay-destination channels are massive multiple-input single-output (MISO) ones. Large random matrix theory (RMT) is used to derive a deterministic equivalent of the signal-to-interference-plus-noise ratio (SINR) at each user in massive MIMO amplify-forward and decode-forward (M-MIMO-ADF) relaying with regularized zero-forcing (RZF) precoding, as the number of transmit antennas and users M,K approaches to infinity and M>>K. In this paper we obtain a closed-form expression for the deterministic equivalent of h^H_kW(hat)_lh(hat)_k, and we give two theorems and a corollary to derive the deterministic equivalent of the SINR at each user. Simulation results show that the deterministic equivalent of the SINR at each user in M-MIMO-ADF relaying and the results of Theorem 1, Theorem 2, Proposition 1 and Corollary 1 are accurate.

eess.SP↗

arXiv2018

An operator that relates to semi-meander polynomials via a two-sided q-Wick formula

We consider the sequence $( Q_n )_{n=1}^{\infty}$ of semi-meander polynomials which are used in the enumeration of semi-meandric systems (a family of diagrams related to the classical stamp-folding problem). We show that for a fixed natural number $d$, the sequence $( Q_n (d) )_{n=1}^{\infty}$ appears as sequence of moments for a compactly supported probability measure $ν_d$ on the real line. More generally, we consider a two-variable generalization $Q_n (t,u)$ of $Q_n(t)$, which is related to a natural concept of "self-intersecting meandric system"; the second variable of $Q_n (t,u)$ keeps track of the crossings of such a system (and one has, in particular, that $Q_n (t,0)$ is the original semi-meander polynomial $Q_n (t)$). We prove that for a fixed natural number $d$ and a fixed real number $q$ with $|q| < 1$, the sequence $( Q_n (d,q) )_{n=1}^{\infty}$ appears as sequence of moments for a compactly supported probability measure $ν_{d:q}$ on the real line. The measure $ν_{d;q}$ is found as scalar spectral measure for an operator $T_{d;q}$ constructed by using left and right creation/annihilation operators on a $q$-deformation of the full Fock space introduced by Bozejko and Speicher. The relevant calculations of moments for $T_{d;q}$ are made by using a two-sided version of a (previously studied in the one-sided case) $q$-Wick formula, which involves the number of crossings of a pair-partition.

math.OA↗

arXiv2017

Free-Boolean independence with amalgamation

In this paper, we develop the notion of free-Boolean independence in an amalgamation setting. We construct free-Boolean cumulants and show that the vanishing of mixed free-Boolean cumulants is equivalent to our free-Boolean independence with amalgamation. We also provide a characterization of free-Boolean independence by conditions in terms of mixed moments. In addition, we study free-Boolean independence over a $C^*$-algebra and prove a positivity property.

math.OA↗

arXiv2017

Continuity and growth of free multiplicative convolution semigroups

Let $μ$ be a compactly supported probability measure on the positive half-line and let $μ^{\boxtimes t}$ be the free multiplicative convolution semigroup. We show that the support of $μ^{\boxtimes t}$ varies continuously as $t$ changes. We also obtain the asymptotic length of the support of these measures.

math.FA↗

arXiv2015

Superconvergence to freely infinitely divisible distributions

The phenomenon of superconvergence is proved for all freely infinitely divisible distributions. Precisely, suppose that the partial sums of a sequence of free identically distributed, infinitesimal random variables converge in distribution to a nondegenerate freely infinitely divisible law. Then the distribution of the sum becomes Lebesgue absolutely continuous with a continuous density in finite time, and this density can be approximated by that of the limit law uniformly, as well as in all $L^{p}$-norms for $p>1$, on the real line except possibly in the neighborhood of one point. Applications include the global superconvergence to freely stable laws and that to free compound Poisson laws over the whole real line.

math.PR↗

arXiv2014

Estimates for compression norms and additivity violation in quantum information

The free contraction norm (or the (t)-norm) was introduced by Belinschi, Collins and Nechita as a tool to compute the typical location of the collection of singular values associated to a random subspace of the tensor product of two Hilbert spaces. In turn, it was used in by them in order to obtain sharp bounds for the violation of the additivity of the minimum output entropy for random quantum channels with Bell states. This free contraction norm, however, is difficult to compute explicitly. The purpose of this note is to give a good estimate for this norm. Our technique is based on results of super convergence in the context of free probability theory. As an application, we give a new, simple and conceptual proof of the violation of the additivity of the minimum output entropy.

math.OA↗

arXiv2014

Local limit theorems for multiplicative free convolutions

This paper describes the quality of convergence to an infinitely divisible law relative to free multiplicative convolution. We show that convergence in distribution for products of identically distributed and infinitesimal free random variables implies superconvergence of their probability densities to the density of the limit law. Superconvergence to the marginal law of free multiplicative Brownian motion at a specified time is also studied. In the unitary case, the superconvergence to free Brownian motion and that to the Haar measure are shown to be uniform over the entire unit circle, implying further a free entropic limit theorem and a universality result for unitary free Lévy processes. Finally, the method of proofs on the positive half-line gives rise to a new multiplicative Boolean to free Bercovici-Pata bijection.

math.FA↗

arXiv2014

On the free convolution with a free multiplicative analogue of the normal distribution

We obtain a formula for the density of the free convolution of an arbitrary probability measure on the unit circle of $\mathbb{C}$ with the free multiplicative analogues of the normal distribution on the unit circle. This description relies on a characterization of the image of the unit disc under the subordination function, which also allows us to prove some regularity properties of the measures obtained in this way. As an application, we give a new proof for Biane's classic result on the densities of the free multiplicative analogue of the normal distributions. We obtain analogue results for probability measures on $\mathbb{R}^+$. Finally, we describe the density of the free multiplicative analogue of the normal distributions as an example and prove unimodality and some symmetry properties of these measures.

math.PR↗