المؤلفون: Liu, Yajing, Zhou, Shijun, Liu, Xiyao, Hao, Chunhui, Fan, Baojie, Tian, Jiandong

مصطلحات موضوعية: Computer Science - Computer Vision and Pattern Recognition

الوصف: Single-source domain generalization (SDG) for object detection is a challenging yet essential task as the distribution bias of the unseen domain degrades the algorithm performance significantly. However, existing methods attempt to extract domain-invariant features, neglecting that the biased data leads the network to learn biased features that are non-causal and poorly generalizable. To this end, we propose an Unbiased Faster R-CNN (UFR) for generalizable feature learning. Specifically, we formulate SDG in object detection from a causal perspective and construct a Structural Causal Model (SCM) to analyze the data bias and feature bias in the task, which are caused by scene confounders and object attribute confounders. Based on the SCM, we design a Global-Local Transformation module for data augmentation, which effectively simulates domain diversity and mitigates the data bias. Additionally, we introduce a Causal Attention Learning module that incorporates a designed attention invariance loss to learn image-level features that are robust to scene confounders. Moreover, we develop a Causal Prototype Learning module with an explicit instance constraint and an implicit prototype constraint, which further alleviates the negative impact of object attribute confounders. Experimental results on five scenes demonstrate the prominent generalization ability of our method, with an improvement of 3.9% mAP on the Night-Clear scene.
Comment: CVPR 2024

الوصول الحر: http://arxiv.org/abs/2405.15225Test

المؤلفون: Yun, Chao, Guo, Haoran, Lin, Zhongchong, Peng, Licong, Liang, Zhongyu, Meng, Miao, Zhang, Biao, Zhao, Zijing, Wang, Leran, Ma, Yifei, Liu, Yajing, Li, Weiwei, Ning, Shuai, Hou, Yanglong, Yang, Jinbo, Luo, Zhaochu

مصطلحات موضوعية: Condensed Matter - Materials Science, Condensed Matter - Mesoscale and Nanoscale Physics

الوصف: The discovery of magnetism in van der Waals (vdW) materials has established unique building blocks for the research of emergent spintronic phenomena. In particular, owing to their intrinsically clean surface without dangling bonds, the vdW magnets hold the potential to construct a superior interface that allows for efficient electrical manipulation of magnetism. Despite several attempts in this direction, it usually requires a cryogenic condition and the assistance of external magnetic fields, which is detrimental to the real application. Here, we fabricate heterostructures based on Fe3GaTe2 flakes that possess room-temperature ferromagnetism with excellent perpendicular magnetic anisotropy. The current-driven non-reciprocal modulation of coercive fields reveals a high spin-torque efficiency in the Fe3GaTe2/Pt heterostructures, which further leads to a full magnetization switching by current. Moreover, we demonstrate the field-free magnetization switching resulting from out-of-plane polarized spin currents by asymmetric geometry design. Our work could expedite the development of efficient vdW spintronic logic, memory and neuromorphic computing devices.

الوصول الحر: http://arxiv.org/abs/2307.01329Test

المؤلفون: Liu, Yajing, Cole, Christina M, Peterson, Chris, Kirby, Michael

مصطلحات موضوعية: Mathematics - Algebraic Topology, Computer Science - Machine Learning, Mathematics - Optimization and Control

الوصف: A ReLU neural network leads to a finite polyhedral decomposition of input space and a corresponding finite dual graph. We show that while this dual graph is a coarse quantization of input space, it is sufficiently robust that it can be combined with persistent homology to detect homological signals of manifolds in the input space from samples. This property holds for a variety of networks trained for a wide range of purposes that have nothing to do with this topological application. We found this feature to be surprising and interesting; we hope it will also be useful.
Comment: Accepted by Proceedings of the 2 nd Annual Workshop on Topology, Algebra, and Geometry in Machine Learning (TAG-ML) at the 40 th In- ternational Conference on Machine Learning

الوصول الحر: http://arxiv.org/abs/2306.17418Test

المؤلفون: Jamil, Huma, Liu, Yajing, Caglar, Turgay, Cole, Christina M., Blanchard, Nathaniel, Peterson, Christopher, Kirby, Michael

مصطلحات موضوعية: Computer Science - Computer Vision and Pattern Recognition, Mathematics - Optimization and Control

الوصف: Researchers typically investigate neural network representations by examining activation outputs for one or more layers of a network. Here, we investigate the potential for ReLU activation patterns (encoded as bit vectors) to aid in understanding and interpreting the behavior of neural networks. We utilize Representational Dissimilarity Matrices (RDMs) to investigate the coherence of data within the embedding spaces of a deep neural network. From each layer of a network, we extract and utilize bit vectors to construct similarity scores between images. From these similarity scores, we build a similarity matrix for a collection of images drawn from 2 classes. We then apply Fiedler partitioning to the associated Laplacian matrix to separate the classes. Our results indicate, through bit vector representations, that the network continues to refine class detectability with the last ReLU layer achieving better than 95\% separation accuracy. Additionally, we demonstrate that bit vectors aid in adversarial image detection, again achieving over 95\% accuracy in separating adversarial and non-adversarial images using a simple classifier.
Comment: accepted by the Workshop TAG in Pattern Recognition with Applications at the Computer Vision and Pattern Recognition (CVPR) 2023

الوصول الحر: http://arxiv.org/abs/2305.01808Test

المؤلفون: Jamil, Huma, Liu, Yajing, Cole, Christina M., Blanchard, Nathaniel, King, Emily J., Kirby, Michael, Peterson, Christopher

مصطلحات موضوعية: Computer Science - Computer Vision and Pattern Recognition, Computer Science - Cryptography and Security, Computer Science - Machine Learning, 68T01, 51M20, 68R10

الوصف: Previous work has shown that a neural network with the rectified linear unit (ReLU) activation function leads to a convex polyhedral decomposition of the input space. These decompositions can be represented by a dual graph with vertices corresponding to polyhedra and edges corresponding to polyhedra sharing a facet, which is a subgraph of a Hamming graph. This paper illustrates how one can utilize the dual graph to detect and analyze adversarial attacks in the context of digital images. When an image passes through a network containing ReLU nodes, the firing or non-firing at a node can be encoded as a bit ($1$ for ReLU activation, $0$ for ReLU non-activation). The sequence of all bit activations identifies the image with a bit vector, which identifies it with a polyhedron in the decomposition and, in turn, identifies it with a vertex in the dual graph. We identify ReLU bits that are discriminators between non-adversarial and adversarial images and examine how well collections of these discriminators can ensemble vote to build an adversarial image detector. Specifically, we examine the similarities and differences of ReLU bit vectors for adversarial images, and their non-adversarial counterparts, using a pre-trained ResNet-50 architecture. While this paper focuses on adversarial digital images, ResNet-50 architecture, and the ReLU activation function, our methods extend to other network architectures, activation functions, and types of datasets.
Comment: 978-1-6654-8045-1/22/\$31.00 \copyright{}2022 IEEE The 6th Workshop on Graph Techniques for Adversarial Activity Analytics (GTA 2022)

الوصول الحر: http://arxiv.org/abs/2211.13305Test

المؤلفون: Zhou Rui, Liu Yajing, Hu Weiqian, Yang Jing, Lin Bing, Zhang Zhentian, Chen Mingyan, Yi Jingwen, Zhu Cuifeng

المصدر: Acta Biochimica et Biophysica Sinica, Vol 56, Pp 844-856 (2024)

مصطلحات موضوعية: Lycium barbarum polysaccharide, lipid droplets, obesity, SIRT1, ATF6, Biochemistry, QD415-436, Genetics, QH426-470

الوصف: Lipid droplets (LDs) are dynamic organelles that store neutral lipids and are closely linked to obesity. Previous studies have suggested that Lycium barbarum polysaccharide (LBP) supplements can ameliorate obesity, but the underlying mechanisms remain unclear. In this study, we hypothesize that LBP alleviates LD accumulation in adipose tissue (AT) by inhibiting fat-specific protein 27 (Fsp27) through an activating transcription factor-6 (ATF6)/small-molecule sirtuin 1 (SIRT1)-dependent mechanism. LD accumulation in AT is induced in high-fat diet (HFD)-fed mice, and differentiation of 3T3-L1 preadipocytes (PAs) is induced. The ability of LBP to alleviate LD accumulation and the possible underlying mechanism are then investigated both in vivo and in vitro. The influences of LBP on the expressions of LD-associated genes ( ATF6 and Fsp27) are also detected. The results show that HFD and PA differentiation markedly increase LD accumulation in ATs and adipocytes, respectively, and these effects are markedly suppressed by LBP supplementation. Furthermore, LBP significantly activates SIRT1 and decreases ATF6 and Fsp27 expressions. Interestingly, the inhibitory effects of LBP are either abolished or exacerbated when ATF6 is overexpressed or silenced, respectively. Furthermore, SIRT1 level is transcriptionally regulated by LBP through opposite actions mediated by ATF6. Collectively, our findings suggest that LBP supplementation alleviates obesity by ameliorating LD accumulation, which might be partially mediated by an ATF6/SIRT1-dependent mechanism.

وصف الملف: electronic resource

العلاقة: https://doaj.org/toc/1672-9145Test

الوصول الحر: https://doaj.org/article/1ed56194a4524dcda27aaf4bc2b468e7Test

المؤلفون: Erickson, Brittany A, Jiang, Junle, Lambert, Valère, Barbot, Sylvain D, Abdelmeguid, Mohamed, Almquist, Martin, Ampuero, Jean-Paul, Ando, Ryosuke, Cattania, Camilla, Chen, Alexandre, Dal Zilio, Luca, Deng, Shuai, Dunham, Eric M, Elbanna, Ahmed E, Gabriel, Alice-Agnes, Harvey, Tobias W, Huang, Yihe, Kaneko, Yoshihiro, Kozdon, Jeremy E, Lapusta, Nadia, Li, Duo, Li, Meng, Liang, Chao, Liu, Yajing, Ozawa, So, Perez-Silva, Andrea, Pranger, Casper, Segall, Paul, Sun, Yudong, Thakur, Prithvi, Uphoff, Carsten, van Dinther, Ylona, Yang, Yuyun

المصدر: Bulletin of the Seismological Society of America. 113(2)

مصطلحات موضوعية: Earth Sciences, Engineering, Geophysics, Civil Engineering, Geochemistry & Geophysics, Civil engineering

الوصف: ABSTRACT: Numerical modeling of earthquake dynamics and derived insight for seismic hazard relies on credible, reproducible model results. The sequences of earthquakes and aseismic slip (SEAS) initiative has set out to facilitate community code comparisons, and verify and advance the next generation of physics-based earthquake models that reproduce all phases of the seismic cycle. With the goal of advancing SEAS models to robustly incorporate physical and geometrical complexities, here we present code comparison results from two new benchmark problems: BP1-FD considers full elastodynamic effects, and BP3-QD considers dipping fault geometries. Seven and eight modeling groups participated in BP1-FD and BP3-QD, respectively, allowing us to explore these physical ingredients across multiple codes and better understand associated numerical considerations. With new comparison metrics, we find that numerical resolution and computational domain size are critical parameters to obtain matching results. Codes for BP1-FD implement different criteria for switching between quasi-static and dynamic solvers, which require tuning to obtain matching results. In BP3-QD, proper remote boundary conditions consistent with specified rigid body translation are required to obtain matching surface displacements. With these numerical and mathematical issues resolved, we obtain excellent quantitative agreements among codes in earthquake interevent times, event moments, and coseismic slip, with reasonable agreements made in peak slip rates and rupture arrival time. We find that including full inertial effects generates events with larger slip rates and rupture speeds compared to the quasi-dynamic counterpart. For BP3-QD, both dip angle and sense of motion (thrust versus normal faulting) alter ground motion on the hanging and foot walls, and influence event patterns, with some sequences exhibiting similar-size characteristic earthquakes, and others exhibiting different-size events. These findings underscore the importance of considering full elastodynamics and nonvertical dip angles in SEAS models, as both influence short- and long-term earthquake behavior and are relevant to seismic hazard.

الوصول الحر: https://escholarship.org/uc/item/7fp0h41gTest

المؤلفون: Lu, Yuning, Wen, Liangjian, Liu, Jianzhuang, Liu, Yajing, Tian, Xinmei

مصطلحات موضوعية: Computer Science - Computer Vision and Pattern Recognition

الوصف: Existing few-shot learning (FSL) methods rely on training with a large labeled dataset, which prevents them from leveraging abundant unlabeled data. From an information-theoretic perspective, we propose an effective unsupervised FSL method, learning representations with self-supervision. Following the InfoMax principle, our method learns comprehensive representations by capturing the intrinsic structure of the data. Specifically, we maximize the mutual information (MI) of instances and their representations with a low-bias MI estimator to perform self-supervised pre-training. Rather than supervised pre-training focusing on the discriminable features of the seen classes, our self-supervised model has less bias toward the seen classes, resulting in better generalization for unseen classes. We explain that supervised pre-training and self-supervised pre-training are actually maximizing different MI objectives. Extensive experiments are further conducted to analyze their FSL performance with various training settings. Surprisingly, the results show that self-supervised pre-training can outperform supervised pre-training under the appropriate conditions. Compared with state-of-the-art FSL methods, our approach achieves comparable performance on widely used FSL benchmarks without any labels of the base classes.
Comment: ECCV 2022, code: https://github.com/bbbdylan/unisiamTest

الوصول الحر: http://arxiv.org/abs/2207.09176Test

المؤلفون: Yu, Hu, Huang, Jie, Liu, Yajing, Zhu, Qi, Zhou, Man, Zhao, Feng

مصطلحات موضوعية: Computer Science - Computer Vision and Pattern Recognition, Computer Science - Artificial Intelligence

الوصف: Deep learning-based source dehazing methods trained on synthetic datasets have achieved remarkable performance but suffer from dramatic performance degradation on real hazy images due to domain shift. Although certain Domain Adaptation (DA) dehazing methods have been presented, they inevitably require access to the source dataset to reduce the gap between the source synthetic and target real domains. To address these issues, we present a novel Source-Free Unsupervised Domain Adaptation (SFUDA) image dehazing paradigm, in which only a well-trained source model and an unlabeled target real hazy dataset are available. Specifically, we devise the Domain Representation Normalization (DRN) module to make the representation of real hazy domain features match that of the synthetic domain to bridge the gaps. With our plug-and-play DRN module, unlabeled real hazy images can adapt existing well-trained source networks. Besides, the unsupervised losses are applied to guide the learning of the DRN module, which consists of frequency losses and physical prior losses. Frequency losses provide structure and style constraints, while the prior loss explores the inherent statistic property of haze-free images. Equipped with our DRN module and unsupervised loss, existing source dehazing models are able to dehaze unlabeled real hazy images. Extensive experiments on multiple baselines demonstrate the validity and superiority of our method visually and quantitatively.
Comment: Accepted to ACM MM 2022

الوصول الحر: http://arxiv.org/abs/2207.06644Test

المؤلفون: Lu, Yuning, Liu, Jianzhuang, Zhang, Yonggang, Liu, Yajing, Tian, Xinmei

مصطلحات موضوعية: Computer Science - Computer Vision and Pattern Recognition

الوصف: We present prompt distribution learning for effectively adapting a pre-trained vision-language model to address downstream recognition tasks. Our method not only learns low-bias prompts from a few samples but also captures the distribution of diverse prompts to handle the varying visual representations. In this way, we provide high-quality task-related content for facilitating recognition. This prompt distribution learning is realized by an efficient approach that learns the output embeddings of prompts instead of the input embeddings. Thus, we can employ a Gaussian distribution to model them effectively and derive a surrogate loss for efficient training. Extensive experiments on 12 datasets demonstrate that our method consistently and significantly outperforms existing methods. For example, with 1 sample per category, it relatively improves the average result by 9.1% compared to human-crafted prompts.
Comment: Accepted by CVPR 2022

الوصول الحر: http://arxiv.org/abs/2205.03340Test