arXiv subjects

Xin Ma

Publications and source records attributed to Xin Ma.

At least 109 records · Page 6Linked to original sources

arXiv2023

SRRM: Semantic Region Relation Model for Indoor Scene Recognition

Despite the remarkable success of convolutional neural networks in various computer vision tasks, recognizing indoor scenes still presents a significant challenge due to their complex composition. Consequently, effectively leveraging semantic information in the scene has been a key issue in advancing indoor scene recognition. Unfortunately, the accuracy of semantic segmentation has limited the effectiveness of existing approaches for leveraging semantic information. As a result, many of these approaches remain at the stage of auxiliary labeling or co-occurrence statistics, with few exploring the contextual relationships between the semantic elements directly within the scene. In this paper, we propose the Semantic Region Relationship Model (SRRM), which starts directly from the semantic information inside the scene. Specifically, SRRM adopts an adaptive and efficient approach to mitigate the negative impact of semantic ambiguity and then models the semantic region relationship to perform scene recognition. Additionally, to more comprehensively exploit the information contained in the scene, we combine the proposed SRRM with the PlacesCNN module to create the Combined Semantic Region Relation Model (CSRRM), and propose a novel information combining approach to effectively explore the complementary contents between them. CSRRM significantly outperforms the SOTA methods on the MIT Indoor 67, reduced Places365 dataset, and SUN RGB-D without retraining. The code is available at: https://github.com/ChuanxinSong/SRRM

cs.CV↗

arXiv2022

Disjoint Masking with Joint Distillation for Efficient Masked Image Modeling

Masked image modeling (MIM) has shown great promise for self-supervised learning (SSL) yet been criticized for learning inefficiency. We believe the insufficient utilization of training signals should be responsible. To alleviate this issue, we introduce a conceptually simple yet learning-efficient MIM training scheme, termed Disjoint Masking with Joint Distillation (DMJD). For disjoint masking (DM), we sequentially sample multiple masked views per image in a mini-batch with the disjoint regulation to raise the usage of tokens for reconstruction in each image while keeping the masking rate of each view. For joint distillation (JD), we adopt a dual branch architecture to respectively predict invisible (masked) and visible (unmasked) tokens with superior learning targets. Rooting in orthogonal perspectives for training efficiency improvement, DM and JD cooperatively accelerate the training convergence yet not sacrificing the model generalization ability. Concretely, DM can train ViT with half of the effective training epochs (3.7 times less time-consuming) to report competitive performance. With JD, our DMJD clearly improves the linear probing classification accuracy over ConvMAE by 5.8%. On fine-grained downstream tasks like semantic segmentation, object detection, etc., our DMJD also presents superior generalization compared with state-of-the-art SSL methods. The code and model will be made public at https://github.com/mx-mark/DMJD.

cs.CV↗

arXiv2022

Accounting for Temporal Variability in Functional Magnetic Resonance Imaging Improves Prediction of Intelligence

Neuroimaging-based prediction methods for intelligence and cognitive abilities have seen a rapid development in literature. Among different neuroimaging modalities, prediction based on functional connectivity (FC) has shown great promise. Most literature has focused on prediction using static FC, but there are limited investigations on the merits of such analysis compared to prediction based on dynamic FC or region level functional magnetic resonance imaging (fMRI) times series that encode temporal variability. To account for the temporal dynamics in fMRI data, we propose a deep neural network involving bi-directional long short-term memory (bi-LSTM) approach that also incorporates feature selection mechanism. The proposed pipeline is implemented via an efficient GPU computation framework and applied to predict intelligence scores based on region level fMRI time series as well as dynamic FC. We compare the prediction performance for different intelligence measures based on static FC, dynamic FC, and region level time series acquired from the Adolescent Brain Cognitive Development (ABCD) study involving close to 7000 individuals. Our detailed analysis illustrates that static FC consistently has inferior prediction performance compared to region level time series or dynamic FC for unimodal rest and task fMRI experiments, and in almost all cases using a combination of task and rest features. In addition, the proposed bi-LSTM pipeline based on region level time series identifies several shared and differential important brain regions across task and rest fMRI experiments that drive intelligence prediction. A test-retest analysis of the selected features shows strong reliability across cross-validation folds. Given the large sample size from ABCD study, our results provide strong evidence that superior prediction of intelligence can be achieved by accounting for temporal variations in fMRI.

q-bio.NC↗

arXiv2022

High-dimensional Measurement Error Models for Lipschitz Loss

Recently emerging large-scale biomedical data pose exciting opportunities for scientific discoveries. However, the ultrahigh dimensionality and non-negligible measurement errors in the data may create difficulties in estimation. There are limited methods for high-dimensional covariates with measurement error, that usually require knowledge of the noise distribution and focus on linear or generalized linear models. In this work, we develop high-dimensional measurement error models for a class of Lipschitz loss functions that encompasses logistic regression, hinge loss and quantile regression, among others. Our estimator is designed to minimize the $L_1$ norm among all estimators belonging to suitable feasible sets, without requiring any knowledge of the noise distribution. Subsequently, we generalize these estimators to a Lasso analog version that is computationally scalable to higher dimensions. We derive theoretical guarantees in terms of finite sample statistical error bounds and sign consistency, even when the dimensionality increases exponentially with the sample size. Extensive simulation studies demonstrate superior performance compared to existing methods in classification and quantile regression problems. An application to a gender classification task based on brain functional connectivity in the Human Connectome Project data illustrates improved accuracy under our approach, and the ability to reliably identify significant brain connections that drive gender differences.

stat.ME↗

arXiv2022

FedSSO: A Federated Server-Side Second-Order Optimization Algorithm

In this work, we propose FedSSO, a server-side second-order optimization method for federated learning (FL). In contrast to previous works in this direction, we employ a server-side approximation for the Quasi-Newton method without requiring any training data from the clients. In this way, we not only shift the computation burden from clients to server, but also eliminate the additional communication for second-order updates between clients and server entirely. We provide theoretical guarantee for convergence of our novel method, and empirically demonstrate our fast convergence and communication savings in both convex and non-convex settings.

cs.LG↗

arXiv2022

A DNS Tunnel Sliding Window Differential Detection Method Based on Normal Distribution Reasonable Range Filtering

A covert attack method often used by APT organizations is the DNS tunnel, which is used to pass information by constructing C2 networks. And they often use the method of frequently changing domain names and server IP addresses to evade monitoring, which makes it extremely difficult to detect them. However, they carry DNS tunnel information traffic in normal DNS communication, which inevitably brings anomalies in some statistical characteristics of DNS traffic, so that it would provide security personnel with the opportunity to find them. Based on the above considerations, this paper studies the statistical discovery methodology of typical DNS tunnel high-frequency query behavior. Firstly, we analyze the distribution of the DNS domain name length and times and finds that the DNS domain name length and times follow the normal distribution law. Secondly, based on this distribution law, we propose a method for detecting and discovering high-frequency DNS query behaviors of non-single domain names based on the statistical rules of domain name length and frequency and we also give three theorems as theoretical support. Thirdly, we design a sliding window difference scheme based on the above method. Experimental results show that our method has a higher detection rate. At the same time, since our method does not need to construct a data set, it has better practicability in detecting unknown DNS tunnels. This also shows that our detection method based on mathematical models can effectively avoid the dilemma for machine learning methods that must have useful training data sets, and has strong practical significance.

cs.NI↗

arXiv2022

A categorical study on the generalized type semigroup

In this short note, we show that the generalized type semigroup $\CW(X, Γ)$ introduced by the author in \cite{M3} belongs to the category \textnormal{W}. In particular, we demonstrate that $\CW(X, Γ)$ satisfies axioms (W1)-(W4) and (W6). When $X$ is zero-dimensional, we also establish (W5) for the semigroup. This supports the analogy between the generalized type semigroup and pre-completed Cuntz semigroup $W(\cdot)$ for $C^*$-algebras.

math.OA↗

arXiv2022

Resolving subcategories and dimensions in recollements of extriangulated categories

Recently, Wang, Wei and Zhang introduced the notion of recollements of extriangulated categories. In this paper, let $(\mathcal{A},\mathcal{B},\mathcal{C})$ be a recollement of extriangulated categories. We provide some methods to construct resolving subcategories in $(\mathcal{A},\mathcal{B},\mathcal{C})$. As applications of the Auslander-Reiten correspondence, we get the gluing of cotilting modules in a recollement of module categories for artin algebras. We also give some bounds of resolution dimensions of the categories involved in $(\mathcal{A},\mathcal{B},\mathcal{C})$ with respect to resolving subcategories, which generalize some known results.

math.RT↗

arXiv2022

TL-GAN: Improving Traffic Light Recognition via Data Synthesis for Autonomous Driving

Traffic light recognition, as a critical component of the perception module of self-driving vehicles, plays a vital role in the intelligent transportation systems. The prevalent deep learning based traffic light recognition methods heavily hinge on the large quantity and rich diversity of training data. However, it is quite challenging to collect data in various rare scenarios such as flashing, blackout or extreme weather, thus resulting in the imbalanced distribution of training data and consequently the degraded performance in recognizing rare classes. In this paper, we seek to improve traffic light recognition by leveraging data synthesis. Inspired by the generative adversarial networks (GANs), we propose a novel traffic light generation approach TL-GAN to synthesize the data of rare classes to improve traffic light recognition for autonomous driving. TL-GAN disentangles traffic light sequence generation into image synthesis and sequence assembling. In the image synthesis stage, our approach enables conditional generation to allow full control of the color of the generated traffic light images. In the sequence assembling stage, we design the style mixing and adaptive template to synthesize realistic and diverse traffic light sequences. Extensive experiments show that the proposed TL-GAN renders remarkable improvement over the baseline without using the generated data, leading to the state-of-the-art performance in comparison with the competing algorithms that are used for general image synthesis and data imbalance tackling.

cs.CV↗

arXiv2022

Boundary actions of CAT(0) spaces and their $C^*$-algebras

In this paper, we study boundary actions of CAT(0) spaces from a point of view of topological dynamics and $C^*$-algebras. First, we investigate the actions of right-angled Coexter groups and right-angled Artin groups with finite defining graphs on the visual boundaries and the Nevo-Sageev boundaries of their natural assigned CAT(0) cube complexes. In particular, we establish (strongly) pure infiniteness results for reduced crossed product $C^*$-algebras of these actions through investigating the corresponding $\cat$ cube complexes and establishing necessary dynamical properties such as minimality, topological freeness and pure infiniteness of the actions. In addition, we study actions of fundamental groups of graphs of groups on the visual boundaries of their Bass-Serre trees. We show that the existence of repeatable paths essentially implies that the action is $2$-filling, from which, we also obtain a large class of unital Kirchberg algebras. Furthermore, our result also provides a new method in identifying $C^*$-simple generalized Baumslag-Solitar groups. The examples of groups obtained from our method have $n$-paradoxical towers in the sense of \cite{G-G-K-N}. This class particularly contains non-degenerated free products, Baumslag-Solitar groups and fundamental groups of $n$-circles or wedge sums of $n$-circles.

math.OA↗

arXiv2022

Multi-task Learning with High-Dimensional Noisy Images

Recent medical imaging studies have given rise to distinct but inter-related datasets corresponding to multiple experimental tasks or longitudinal visits. Standard scalar-on-image regression models that fit each dataset separately are not equipped to leverage information across inter-related images, and existing multi-task learning approaches are compromised by the inability to account for the noise that is often observed in images. We propose a novel joint scalar-on-image regression framework involving wavelet-based image representations with grouped penalties that are designed to pool information across inter-related images for joint learning, and which explicitly accounts for noise in high-dimensional images via a projection-based approach. In the presence of non-convexity arising due to noisy images, we derive non-asymptotic error bounds under non-convex as well as convex grouped penalties, even when the number of voxels increases exponentially with sample size. A projected gradient descent algorithm is used for computation, which is shown to approximate the optimal solution via well-defined non-asymptotic optimization error bounds under noisy images. Extensive simulations and application to a motivating longitudinal Alzheimer's disease study illustrate significantly improved predictive ability and greater power to detect true signals, that are simply missed by existing methods without noise correction due to the attenuation to null phenomenon.

stat.ME↗

arXiv2022

Compressing Models with Few Samples: Mimicking then Replacing

Few-sample compression aims to compress a big redundant model into a small compact one with only few samples. If we fine-tune models with these limited few samples directly, models will be vulnerable to overfit and learn almost nothing. Hence, previous methods optimize the compressed model layer-by-layer and try to make every layer have the same outputs as the corresponding layer in the teacher model, which is cumbersome. In this paper, we propose a new framework named Mimicking then Replacing (MiR) for few-sample compression, which firstly urges the pruned model to output the same features as the teacher's in the penultimate layer, and then replaces teacher's layers before penultimate with a well-tuned compact one. Unlike previous layer-wise reconstruction methods, our MiR optimizes the entire network holistically, which is not only simple and effective, but also unsupervised and general. MiR outperforms previous methods with large margins. Codes will be available soon.

cs.LG↗

arXiv2021

Contrastive Attention Network with Dense Field Estimation for Face Completion

Most modern face completion approaches adopt an autoencoder or its variants to restore missing regions in face images. Encoders are often utilized to learn powerful representations that play an important role in meeting the challenges of sophisticated learning tasks. Specifically, various kinds of masks are often presented in face images in the wild, forming complex patterns, especially in this hard period of COVID-19. It's difficult for encoders to capture such powerful representations under this complex situation. To address this challenge, we propose a self-supervised Siamese inference network to improve the generalization and robustness of encoders. It can encode contextual semantics from full-resolution images and obtain more discriminative representations. To deal with geometric variations of face images, a dense correspondence field is integrated into the network. We further propose a multi-scale decoder with a novel dual attention fusion module (DAF), which can combine the restored and known regions in an adaptive manner. This multi-scale architecture is beneficial for the decoder to utilize discriminative representations learned from encoders into images. Extensive experiments clearly demonstrate that the proposed approach not only achieves more appealing results compared with state-of-the-art methods but also improves the performance of masked face recognition dramatically.

cs.CV↗

arXiv2021

Fiberwise amenability of ample étale groupoids

Let $\mathcal{G}$ be a locally compact $σ$-compact Hausdorff ample groupoid on a compact space. In this paper, we further examine the (ubiquitous) fiberwise amenability introduced by the author and Jianchao Wu for $\mathcal{G}$. We define the corresponding concepts of Følner sequences and Banach densities for $\mathcal{G}$, based on which, we establish a topological groupoid version of the Ornstein-Weiss quasi-tilling theorem. This leads to the notion of almost finiteness in measure for ample groupoids as a weaker version of Matui's almost finiteness. As applications, we first show that $C^*_r(\mathcal{G})$ has the uniform property $Γ$ and thus satisfies the Toms-Winter conjecture when $\mathcal{G}$ is minimal second countable (topologically) amenable and almost finite in measure. Then we prove that the topological full group $[[\mathcal{G}]]$ is always sofic when $\mathcal{G}$ is second countable minimal and admits a Følner sequence. This can be used to strengthen one of Matui's result on the commutator subgroup $D[[\mathcal{G}]]$ when $\mathcal{G}$ is almost finite. Concrete examples are provided.

math.OA↗

arXiv2021

Separating Boundary Points via Structural Regularization for Very Compact Clusters

Clustering algorithms have significantly improved along with Deep Neural Networks which provide effective representation of data. Existing methods are built upon deep autoencoder and self-training process that leverages the distribution of cluster assignments of samples. However, as the fundamental objective of the autoencoder is focused on efficient data reconstruction, the learnt space may be sub-optimal for clustering. Moreover, it requires highly effective codes (i.e., representation) of data, otherwise the initial cluster centers often cause stability issues during self-training. Many state-of-the-art clustering algorithms use convolution operation to extract efficient codes but their applications are limited to image data. In this regard, we propose an end-to-end deep clustering algorithm, i.e., Very Compact Clusters (VCC). VCC takes advantage of distributions of local relationships of samples near the boundary of clusters, so that they can be properly separated and pulled to cluster centers to form compact clusters. Experimental results on various datasets illustrate that our proposed approach achieves competitive clustering performance against most of the state-of-the-art clustering methods for both image and non-image data, and its results can be easily qualitatively seen in the learnt low-dimensional space.

cs.CV↗

arXiv2021

Interfacial Dzyaloshinskii-Moriya interaction and spin-orbit torque in Au1-xPtx/Co bilayers with varying interfacial spin-orbit coupling

The quantitative roles of the interfacial spin-orbit coupling (SOC) in Dzyaloshinskii-Moriya interaction (DMI) and dampinglike spin-orbit torque (τDL) have remained unsettled after a decade of intensive study. Here, we report a conclusive experiment evidence that, because of the critical role of the interfacial orbital hybridization, the interfacial DMI is not necessarily a linear function of the interfacial SOC, e.g. at Au1-xPtx/Co interfaces where the interfacial SOC can be tuned significantly via strongly composition (x)-dependent spin-orbit proximity effect without varying the bulk SOC and the electronegativity of the Au1-xPtx layer. We also find that τDL in the Au1-xPtx/Co bilayers varies distinctly from the interfacial SOC as a function of x, indicating no important τDL contribution from the interfacial Rashba-Edelstein effect.

cond-mat.mtrl-sci↗

arXiv2021

Learning Multimodal Contact-Rich Skills from Demonstrations Without Reward Engineering

Everyday contact-rich tasks, such as peeling, cleaning, and writing, demand multimodal perception for effective and precise task execution. However, these present a novel challenge to robots as they lack the ability to combine these multimodal stimuli for performing contact-rich tasks. Learning-based methods have attempted to model multi-modal contact-rich tasks, but they often require extensive training examples and task-specific reward functions which limits their practicality and scope. Hence, we propose a generalizable model-free learning-from-demonstration framework for robots to learn contact-rich skills without explicit reward engineering. We present a novel multi-modal sensor data representation which improves the learning performance for contact-rich skills. We performed training and experiments using the real-life Sawyer robot for three everyday contact-rich skills -- cleaning, writing, and peeling. Notably, the framework achieves a success rate of 100\% for the peeling and writing skill, and 80\% for the cleaning skill. Hence, this skill learning framework can be extended for learning other physical manipulation skills.

cs.RO↗

arXiv2021

Homological Dimensions of Extriangulated Categories and Recollements

In this paper, let $(\mathcal{A},\mathcal{B},\mathcal{C})$ be a recollement of extriangulated categories. We introduce the global dimension and extension dimension of extriangulated categories, and give some upper bounds of global dimensions (resp. extension dimensions) of the categories involved in $(\mathcal{A},\mathcal{B},\mathcal{C})$, which give a simultaneous generalization of some results in the recollement of abelian categories and triangulated categories.

math.RT↗