Search

Recent work by Ram\'irez et al. [2] has introduced Multi-Channel Factor Analysis (MFA) as an extension of factor analysis to multi-channel data that allows for latent factors common to all channels as well as factors specific to each channel. This paper validates the MFA covariance model and analyzes the statistical properties of the MFA estimators. In particular, a thorough investigation of model identifiability under varying latent factor structures is conducted, and sufficient conditions for generic global identifiability of MFA are obtained. The development of these identifiability conditions enables asymptotic analysis of estimators obtained by maximizing a Gaussian likelihood, which are shown to be consistent and asymptotically normal even under misspecification of the latent factor distribution.

Many fascinating properties discovered in graphene and transition metal dichalcogenide (TMD) moir\'e superlattices originate from flat bands and enhanced many-body effects. Here, we discover new many-electron excited states in TMD homobilayers. As optical resonances evolve with twist angle and doping in MoSe$_2$ bilayers, a unique type of ``charge-transfer" trions is observed when gradual changes in atomic alignment between the layers occur. In real space, the optically excited electron-hole pair mostly resides in a different site from the doped hole in a moir\'e supercell. In momentum space, the electron-hole pair forms in the single-particle-band $K$-valley, while the hole occupies the $\Gamma$-valley. The rich internal structure of this trion resonance arises from the ultra-flatness of the first valence band and the distinct influence of moir\'e potential modulation on holes and excitons. Our findings open new routes to realizing photon-spin transduction or implementing moir\'e quantum simulators with independently tunable fermion and boson densities., Comment: 11 pages, 10 figures

In recent years, 2D human pose estimation has made significant progress on public benchmarks. However, many of these approaches face challenges of less applicability in the industrial community due to the large number of parametric quantities and computational overhead. Efficient human pose estimation remains a hurdle, especially for whole-body pose estimation with numerous keypoints. While most current methods for efficient human pose estimation primarily rely on CNNs, we propose the Group-based Token Pruning Transformer (GTPT) that fully harnesses the advantages of the Transformer. GTPT alleviates the computational burden by gradually introducing keypoints in a coarse-to-fine manner. It minimizes the computation overhead while ensuring high performance. Besides, GTPT groups keypoint tokens and prunes visual tokens to improve model performance while reducing redundancy. We propose the Multi-Head Group Attention (MHGA) between different groups to achieve global interaction with little computational overhead. We conducted experiments on COCO and COCO-WholeBody. Compared to other methods, the experimental results show that GTPT can achieve higher performance with less computation, especially in whole-body with numerous keypoints., Comment: ECCV 2024 accepted

Semi-supervised medical image segmentation (SSMIS) has emerged as a promising solution to tackle the challenges of time-consuming manual labeling in the medical field. However, in practical scenarios, there are often domain variations within the datasets, leading to derivative scenarios like semi-supervised medical domain generalization (Semi-MDG) and unsupervised medical domain adaptation (UMDA). In this paper, we aim to develop a generic framework that masters all three tasks. We notice a critical shared challenge across three scenarios: the explicit semantic knowledge for segmentation performance and rich domain knowledge for generalizability exclusively exist in the labeled set and unlabeled set respectively. Such discrepancy hinders existing methods from effectively comprehending both types of knowledge under semi-supervised settings. To tackle this challenge, we develop a Semantic & Domain Knowledge Messenger (S&D Messenger) which facilitates direct knowledge delivery between the labeled and unlabeled set, and thus allowing the model to comprehend both of them in each individual learning flow. Equipped with our S&D Messenger, a naive pseudo-labeling method can achieve huge improvement on six benchmark datasets for SSMIS (+7.5%), UMDA (+5.6%), and Semi-MDG tasks (+1.14%), compared with state-of-the-art methods designed for specific tasks., Comment: 10 pages, under review of IEEE Transcations on Medical Imaging

Bi-level optimization (BO) has become a fundamental mathematical framework for addressing hierarchical machine learning problems. As deep learning models continue to grow in size, the demand for scalable bi-level optimization solutions has become increasingly critical. Traditional gradient-based bi-level optimization algorithms, due to their inherent characteristics, are ill-suited to meet the demands of large-scale applications. In this paper, we introduce $\textbf{F}$orward $\textbf{G}$radient $\textbf{U}$nrolling with $\textbf{F}$orward $\textbf{F}$radient, abbreviated as $(\textbf{FG})^2\textbf{U}$, which achieves an unbiased stochastic approximation of the meta gradient for bi-level optimization. $(\text{FG})^2\text{U}$ circumvents the memory and approximation issues associated with classical bi-level optimization approaches, and delivers significantly more accurate gradient estimates than existing large-scale bi-level optimization approaches. Additionally, $(\text{FG})^2\text{U}$ is inherently designed to support parallel computing, enabling it to effectively leverage large-scale distributed computing systems to achieve significant computational efficiency. In practice, $(\text{FG})^2\text{U}$ and other methods can be strategically placed at different stages of the training process to achieve a more cost-effective two-phase paradigm. Further, $(\text{FG})^2\text{U}$ is easy to implement within popular deep learning frameworks, and can be conveniently adapted to address more challenging zeroth-order bi-level optimization scenarios. We provide a thorough convergence analysis and a comprehensive practical discussion for $(\text{FG})^2\text{U}$, complemented by extensive empirical evaluations, showcasing its superior performance in diverse large-scale bi-level optimization tasks.

The ferrimagnet TbMn6Sn6 has attracted vast attention, because its pristine Mn kagome lattice with strong spin-orbit coupling and out-of-plane Tb-Mn exchange supports quantum-limit Chern topological magnetism which can be described by the simple spinless Haldane model. We unveil herein that engineering the pristine kagome lattice through partial replacement of Mn by nonmagnetic Cr which tends to concentrate into the single Mn1 layer in a unit cell breaks the collinear configuration of Mn spins and reduces the D6h point group symmetry to the C2 one. The nearly isolated Tb networks result in easily polarized Tb spins even under a weak magnetic field, and simultaneously, different spin chirality of the Tb-Mn1-Mn1 and Mn1-Mn1-Mn1. Such a peculiar spin structure leads to a plateau-like topological Hall effect with a record resistivity of 19.1 {\mu}Ohm cm among bulk systems. Our direct visualization of the domain-wall structure and its evolution under external magnetic field fully support the picture, thus highlighting the pivotal role of broken kagome lattice symmetry in generating the peculiar spin chirality in real space. Our results set a paradigm for exploration of exotic properties in kagome topological magnets and would be a proof-of-principle strategy for investigating the correlation between magnetism and exotic topological properties in kagome lattice., Comment: 33 pages,4 main figures and 16 SI figures

Impurity substitution is a general strategy to study the intrinsic properties of a quantum material. However, when the target element has more than one Wyckoff position in the lattice, it is a big challenge but with extreme necessity to know the exact position and order of the occupancy of impurity atoms. Via comprehensive experimental and theoretical investigations, we establish herein the roadmap for Ni substitution in Fe3GaTe2, a van der Waals ferromagnet with the Curie temperature TC even reaching ~ 380 K. The results unambiguously reveal that in (Fe1-xNix)3GaTe2, Ni atoms initially form an van der Waals interlayer gap Ni3 sites when x < 0.1, and then gradually occupy the Fe2 sites. After replacing the Fe2 sites at x of ~ 0.75, they start to substitute for the Fe1 sites and eventually realize a full occupation at x = 1.0. Accordingly, TC and saturation magnetic moments of (Fe1-xNix)3GaTe2 both show nonlinear decrease, which is tightly tied to the Ni occupancy order as well as the different roles of Ni3, Fe1 and Fe2 sites in the spin Hamiltonian. The results not only yield fruitful insights into the essential roles of different Fe sites in producing the above room temperature high TC, but also set a paradigm for future impurity substitution study on other quantum materials., Comment: 24 pages,5 main figures+4 SI figures+2 SI tables

This study introduces a novel data augmentation technique, ADLDA, aimed at mitigating the negative impact of data distribution shifts caused by the data augmentation process in computer vision task. ADLDA partitions augmented data into distinct subdomains and incorporates domain labels, combined with domain adaptation techniques, to optimize data representation in the model's feature space. Experimental results demonstrate that ADLDA significantly enhances model performance across multiple datasets, particularly in neural network architectures with complex feature extraction layers. Furthermore, ADLDA improves the model's ability to locate and recognize key features, showcasing potential in object recognition and image segmentation tasks. This paper's contribution provides an effective data augmentation regularization method for the field of computer vision aiding in the enhancement of robustness and accuracy in deep learning models., Comment: 8 page 4 fig

Owing to the outstanding properties provided by nontrivial band topology, topological phases of matter are considered as a promising platform towards low-dissipation electronics, efficient spin-charge conversion, and topological quantum computation. Achieving ferroelectricity in topological materials enables the non-volatile control of the quantum states, which could greatly facilitate topological electronic research. However, ferroelectricity is generally incompatible with systems featuring metallicity due to the screening effect of free carriers. In this study, we report the observation of memristive switching based on the ferroelectric surface state of a topological semimetal (TaSe4)2I. We find that the surface state of (TaSe4)2I presents out-of-plane ferroelectric polarization due to surface reconstruction. With the combination of ferroelectric surface and charge-density-wave-gapped bulk states, an electric switchable barrier height can be achieved in (TaSe4)2I-metal contact. By employing a multi-terminal grounding design, we manage to construct a prototype ferroelectric memristor based on (TaSe4)2I with on/off ratio up to 10^3, endurance over 10^3 cycles, and good retention characteristics. The origin of the ferroelectric surface state is further investigated by first-principles calculations, which reveals an interplay between ferroelectricity and band topology. The emergence of ferroelectricity in (TaSe4)2I not only demonstrates it as a rare but essential case of ferroelectric topological materials, but also opens new routes towards the implementation of topological materials in functional electronic devices., Comment: 14 pages, 5 figures

In recent years, with the rapid development of computer information technology, the development of artificial intelligence has been accelerating. The traditional geometry recognition technology is relatively backward and the recognition rate is low. In the face of massive information database, the traditional algorithm model inevitably has the problems of low recognition accuracy and poor performance. Deep learning theory has gradually become a very important part of machine learning. The implementation of convolutional neural network (CNN) reduces the difficulty of graphics generation algorithm. In this paper, using the advantages of lenet-5 architecture sharing weights and feature extraction and classification, the proposed geometric pattern recognition algorithm model is faster in the training data set. By constructing the shared feature parameters of the algorithm model, the cross-entropy loss function is used in the recognition process to improve the generalization of the model and improve the average recognition accuracy of the test data set., Comment: 6 pages,9 figures

In text-to-image generation tasks, the advancements of diffusion models have facilitated the fidelity of generated results. However, these models encounter challenges when processing text prompts containing multiple entities and attributes. The uneven distribution of attention results in the issues of entity leakage and attribute misalignment. Training from scratch to address this issue requires numerous labeled data and is resource-consuming. Motivated by this, we propose an attribution-focusing mechanism, a training-free phase-wise mechanism by modulation of attention for diffusion model. One of our core ideas is to guide the model to concentrate on the corresponding syntactic components of the prompt at distinct timesteps. To achieve this, we incorporate a temperature control mechanism within the early phases of the self-attention modules to mitigate entity leakage issues. An object-focused masking scheme and a phase-wise dynamic weight control mechanism are integrated into the cross-attention modules, enabling the model to discern the affiliation of semantic information between entities more effectively. The experimental results in various alignment scenarios demonstrate that our model attain better image-text alignment with minimal additional computational cost.

We develop theory leading to testing procedures for the presence of a change point in the intraday volatility pattern. The new theory is developed in the framework of Functional Data Analysis. It is based on a model akin to the stochastic volatility model for scalar point-to-point returns. In our context, we study intraday curves, one curve per trading day. After postulating a suitable model for such functional data, we present three tests focusing, respectively, on changes in the shape, the magnitude and arbitrary changes in the sequences of the curves of interest. We justify the respective procedures by showing that they have asymptotically correct size and by deriving consistency rates for all tests. These rates involve the sample size (the number of trading days) and the grid size (the number of observations per day). We also derive the corresponding change point estimators and their consistency rates. All procedures are additionally validated by a simulation study and an application to US stocks.

Semi-supervised semantic segmentation (SSSS) has been proposed to alleviate the burden of time-consuming pixel-level manual labeling, which leverages limited labeled data along with larger amounts of unlabeled data. Current state-of-the-art methods train the labeled data with ground truths and unlabeled data with pseudo labels. However, the two training flows are separate, which allows labeled data to dominate the training process, resulting in low-quality pseudo labels and, consequently, sub-optimal results. To alleviate this issue, we present AllSpark, which reborns the labeled features from unlabeled ones with the channel-wise cross-attention mechanism. We further introduce a Semantic Memory along with a Channel Semantic Grouping strategy to ensure that unlabeled features adequately represent labeled features. The AllSpark shed new light on the architecture level designs of SSSS rather than framework level, which avoids increasingly complicated training pipeline designs. It can also be regarded as a flexible bottleneck module that can be seamlessly integrated into a general transformer-based segmentation model. The proposed AllSpark outperforms existing methods across all evaluation protocols on Pascal, Cityscapes and COCO benchmarks without bells-and-whistles. Code and model weights are available at: https://github.com/xmed-lab/AllSpark., Comment: Accepted by CVPR 2024; correct typos; this is not the camera-ready version

The commercialization of text-to-image diffusion models (DMs) brings forth potential copyright concerns. Despite numerous attempts to protect DMs from copyright issues, the vulnerabilities of these solutions are underexplored. In this study, we formalized the Copyright Infringement Attack on generative AI models and proposed a backdoor attack method, SilentBadDiffusion, to induce copyright infringement without requiring access to or control over training processes. Our method strategically embeds connections between pieces of copyrighted information and text references in poisoning data while carefully dispersing that information, making the poisoning data inconspicuous when integrated into a clean dataset. Our experiments show the stealth and efficacy of the poisoning data. When given specific text prompts, DMs trained with a poisoning ratio of 0.20% can produce copyrighted images. Additionally, the results reveal that the more sophisticated the DMs are, the easier the success of the attack becomes. These findings underline potential pitfalls in the prevailing copyright protection strategies and underscore the necessity for increased scrutiny to prevent the misuse of DMs., Comment: Accepted for presentation at ICML 2024

Creativity serves as a cornerstone for societal progress and innovation. With the rise of advanced generative AI models capable of tasks once reserved for human creativity, the study of AI's creative potential becomes imperative for its responsible development and application. In this paper, we prove in theory that AI can be as creative as humans under the condition that it can properly fit the data generated by human creators. Therefore, the debate on AI's creativity is reduced into the question of its ability to fit a sufficient amount of data. To arrive at this conclusion, this paper first addresses the complexities in defining creativity by introducing a new concept called Relative Creativity. Rather than attempting to define creativity universally, we shift the focus to whether AI can match the creative abilities of a hypothetical human. The methodological shift leads to a statistically quantifiable assessment of AI's creativity, term Statistical Creativity. This concept, statistically comparing the creative abilities of AI with those of specific human groups, facilitates theoretical exploration of AI's creative potential. Our analysis reveals that by fitting extensive conditional data without marginalizing out the generative conditions, AI can emerge as a hypothetical new creator. The creator possesses the same creative abilities on par with the human creators it was trained on. Building on theoretical findings, we discuss the application in prompt-conditioned autoregressive models, providing a practical means for evaluating creative abilities of generative AI models, such as Large Language Models (LLMs). Additionally, this study provides an actionable training guideline, bridging the theoretical quantification of creativity with practical model training., Comment: The paper examines AI's creativity, introducing Relative and Statistical Creativity for theoretical and practical analysis, along with practical training guidelines. Project Page: ai-relative-creativity.github.io

Most state-of-the-art methods for medical image segmentation adopt the encoder-decoder architecture. However, this U-shaped framework still has limitations in capturing the non-local multi-scale information with a simple skip connection. To solve the problem, we firstly explore the potential weakness of skip connections in U-Net on multiple segmentation tasks, and find that i) not all skip connections are useful, each skip connection has different contribution; ii) the optimal combinations of skip connections are different, relying on the specific datasets. Based on our findings, we propose a new segmentation framework, named UDTransNet, to solve three semantic gaps in U-Net. Specifically, we propose a Dual Attention Transformer (DAT) module for capturing the channel- and spatial-wise relationships to better fuse the encoder features, and a Decoder-guided Recalibration Attention (DRA) module for effectively connecting the DAT tokens and the decoder features to eliminate the inconsistency. Hence, both modules establish a learnable connection to solve the semantic gaps between the encoder and the decoder, which leads to a high-performance segmentation model for medical images. Comprehensive experimental results indicate that our UDTransNet produces higher evaluation scores and finer segmentation results with relatively fewer parameters over the state-of-the-art segmentation methods on different public datasets. Code: https://github.com/McGregorWwww/UDTransNet.

Dataset distillation reduces the storage and computational consumption of training a network by generating a small surrogate dataset that encapsulates rich information of the original large-scale one. However, previous distillation methods heavily rely on the sample-wise iterative optimization scheme. As the images-per-class (IPC) setting or image resolution grows larger, the necessary computation will demand overwhelming time and resources. In this work, we intend to incorporate generative diffusion techniques for computing the surrogate dataset. Observing that key factors for constructing an effective surrogate dataset are representativeness and diversity, we design additional minimax criteria in the generative training to enhance these facets for the generated images of diffusion models. We present a theoretical model of the process as hierarchical diffusion control demonstrating the flexibility of the diffusion process to target these criteria without jeopardizing the faithfulness of the sample to the desired distribution. The proposed method achieves state-of-the-art validation performance while demanding much less computational resources. Under the 100-IPC setting on ImageWoof, our method requires less than one-twentieth the distillation time of previous methods, yet yields even better performance. Source code and generated data are available in https://github.com/vimar-gu/MinimaxDiffusion., Comment: CVPR 2024

Volume-wise labeling in 3D medical images is a time-consuming task that requires expertise. As a result, there is growing interest in using semi-supervised learning (SSL) techniques to train models with limited labeled data. However, the challenges and practical applications extend beyond SSL to settings such as unsupervised domain adaptation (UDA) and semi-supervised domain generalization (SemiDG). This work aims to develop a generic SSL framework that can handle all three settings. We identify two main obstacles to achieving this goal in the existing SSL framework: 1) the weakness of capturing distribution-invariant features; and 2) the tendency for unlabeled data to be overwhelmed by labeled data, leading to over-fitting to the labeled data during training. To address these issues, we propose an Aggregating & Decoupling framework. The aggregating part consists of a Diffusion encoder that constructs a common knowledge set by extracting distribution-invariant features from aggregated information from multiple distributions/domains. The decoupling part consists of three decoders that decouple the training process with labeled and unlabeled data, thus avoiding over-fitting to labeled data, specific domains and classes. We evaluate our proposed framework on four benchmark datasets for SSL, Class-imbalanced SSL, UDA and SemiDG. The results showcase notable improvements compared to state-of-the-art methods across all four settings, indicating the potential of our framework to tackle more challenging SSL scenarios. Code and models are available at: https://github.com/xmed-lab/GenericSSL., Comment: Accepted at NeurIPS 2023

Embedding-based retrieval methods construct vector indices to search for document representations that are most similar to the query representations. They are widely used in document retrieval due to low latency and decent recall performance. Recent research indicates that deep retrieval solutions offer better model quality, but are hindered by unacceptable serving latency and the inability to support document updates. In this paper, we aim to enhance the vector index with end-to-end deep generative models, leveraging the differentiable advantages of deep retrieval models while maintaining desirable serving efficiency. We propose Model-enhanced Vector Index (MEVI), a differentiable model-enhanced index empowered by a twin-tower representation model. MEVI leverages a Residual Quantization (RQ) codebook to bridge the sequence-to-sequence deep retrieval and embedding-based models. To substantially reduce the inference time, instead of decoding the unique document ids in long sequential steps, we first generate some semantic virtual cluster ids of candidate documents in a small number of steps, and then leverage the well-adapted embedding vectors to further perform a fine-grained search for the relevant documents in the candidate virtual clusters. We empirically show that our model achieves better performance on the commonly used academic benchmarks MSMARCO Passage and Natural Questions, with comparable serving latency to dense retrieval solutions.

Diffusion models excel in many generative modeling tasks, notably in creating images from text prompts, a task referred to as text-to-image (T2I) generation. Despite the ability to generate high-quality images, these models often replicate elements from their training data, leading to increasing copyright concerns in real applications in recent years. In response to this raising concern about copyright infringement, recent studies have studied the copyright behavior of diffusion models when using direct, copyrighted prompts. Our research extends this by examining subtler forms of infringement, where even indirect prompts can trigger copyright issues. Specifically, we introduce a data generation pipeline to systematically produce data for studying copyright in diffusion models. Our pipeline enables us to investigate copyright infringement in a more practical setting, involving replicating visual features rather than entire works using seemingly irrelevant prompts for T2I generation. We generate data using our proposed pipeline to test various diffusion models, including the latest Stable Diffusion XL. Our findings reveal a widespread tendency that these models tend to produce copyright-infringing content, highlighting a significant challenge in this field., Comment: 16 pages including appendix

Moir\'e potential acts as periodic quantum confinement for optically generated exciton, generating spatially ordered zero-dimensional quantum system. However, broad emission spectrum arising from inhomogeneity among moir\'e potential hinders the exploration of the intrinsic properties of moir\'e exciton. In this study, we have demonstrated a new method to realize the optical observation of quantum coherence and interference of a single moir\'e exciton in twisted semiconducting heterobilayer beyond the diffraction limit of light. A significant single and sharp photoluminescence peak from a single moir\'e exciton has been demonstrated after nano-fabrication. We present the longer duration of quantum coherence of a single moir\'e exciton, which reaches beyond 10 ps and the accelerated decoherence process with elevating temperature and excitation power density. Moreover, the quantum interference has revealed the coupling between moir\'e excitons in different moir\'e potential minima. The observed quantum coherence and interference of moir\'e exciton will facilitate potential application toward quantum technologies based on moir\'e quantum systems., Comment: 42 pages, 4 figures

Modern dataset search platforms employ ML task-based utility metrics instead of relying on metadata-based keywords to comb through extensive dataset repositories. In this setup, requesters provide an initial dataset, and the platform identifies complementary datasets to augment (join or union) the requester's dataset such that the ML model (e.g., linear regression) performance is improved most. Although effective, current task-based data searches are stymied by (1) high latency which deters users, (2) privacy concerns for regulatory standards, and (3) low data quality which provides low utility. We introduce Mileena, a fast, private, and high-quality task-based dataset search platform. At its heart, Mileena is built on pre-computed semi-ring sketches for efficient ML training and evaluation. Based on semi-ring, we develop a novel Factorized Privacy Mechanism that makes the search differentially private and scales to arbitrary corpus sizes and numbers of requests without major quality degradation. We also demonstrate the early promise in using LLM-based agents for automatic data transformation and applying semi-rings to support causal discovery and treatment effect estimation.

Heatmap-based methods have become the mainstream method for pose estimation due to their superior performance. However, heatmap-based approaches suffer from significant quantization errors with downscale heatmaps, which result in limited performance and the detrimental effects of intermediate supervision. Previous heatmap-based methods relied heavily on additional post-processing to mitigate quantization errors. Some heatmap-based approaches improve the resolution of feature maps by using multiple costly upsampling layers to improve localization precision. To solve the above issues, we creatively view the backbone network as a degradation process and thus reformulate the heatmap prediction as a Super-Resolution (SR) task. We first propose the SR head, which predicts heatmaps with a spatial resolution higher than the input feature maps (or even consistent with the input image) by super-resolution, to effectively reduce the quantization error and the dependence on further post-processing. Besides, we propose SRPose to gradually recover the HR heatmaps from LR heatmaps and degraded features in a coarse-to-fine manner. To reduce the training difficulty of HR heatmaps, SRPose applies SR heads to supervise the intermediate features in each stage. In addition, the SR head is a lightweight and generic head that applies to top-down and bottom-up methods. Extensive experiments on the COCO, MPII, and CrowdPose datasets show that SRPose outperforms the corresponding heatmap-based approaches. The code and models are available at https://github.com/haonanwang0522/SRPose., Comment: ACM MM 2023 accepted

The volume-wise labeling of 3D medical images is expertise-demanded and time-consuming; hence semi-supervised learning (SSL) is highly desirable for training with limited labeled data. Imbalanced class distribution is a severe problem that bottlenecks the real-world application of these methods but was not addressed much. Aiming to solve this issue, we present a novel Dual-debiased Heterogeneous Co-training (DHC) framework for semi-supervised 3D medical image segmentation. Specifically, we propose two loss weighting strategies, namely Distribution-aware Debiased Weighting (DistDW) and Difficulty-aware Debiased Weighting (DiffDW), which leverage the pseudo labels dynamically to guide the model to solve data and learning biases. The framework improves significantly by co-training these two diverse and accurate sub-models. We also introduce more representative benchmarks for class-imbalanced semi-supervised medical image segmentation, which can fully demonstrate the efficacy of the class-imbalance designs. Experiments show that our proposed framework brings significant improvements by using pseudo labels for debiasing and alleviating the class imbalance problem. More importantly, our method outperforms the state-of-the-art SSL methods, demonstrating the potential of our framework for the more challenging SSL setting. Code and models are available at: https://github.com/xmed-lab/DHC., Comment: Accepted at MICCAI2023

In generative modeling, numerous successful approaches leverage a low-dimensional latent space, e.g., Stable Diffusion models the latent space induced by an encoder and generates images through a paired decoder. Although the selection of the latent space is empirically pivotal, determining the optimal choice and the process of identifying it remain unclear. In this study, we aim to shed light on this under-explored topic by rethinking the latent space from the perspective of model complexity. Our investigation starts with the classic generative adversarial networks (GANs). Inspired by the GAN training objective, we propose a novel "distance" between the latent and data distributions, whose minimization coincides with that of the generator complexity. The minimizer of this distance is characterized as the optimal data-dependent latent that most effectively capitalizes on the generator's capacity. Then, we consider parameterizing such a latent distribution by an encoder network and propose a two-stage training strategy called Decoupled Autoencoder (DAE), where the encoder is only updated in the first stage with an auxiliary decoder and then frozen in the second stage while the actual decoder is being trained. DAE can improve the latent distribution and as a result, improve the generative performance. Our theoretical analyses are corroborated by comprehensive experiments on various models such as VQGAN and Diffusion Transformer, where our modifications yield significant improvements in sample quality with decreased model complexity., Comment: Accepted to NeurIPS 2023 (Spotlight)

Long-tailed recognition with imbalanced class distribution naturally emerges in practical machine learning applications. Existing methods such as data reweighing, resampling, and supervised contrastive learning enforce the class balance with a price of introducing imbalance between instances of head class and tail class, which may ignore the underlying rich semantic substructures of the former and exaggerate the biases in the latter. We overcome these drawbacks by a novel ``subclass-balancing contrastive learning (SBCL)'' approach that clusters each head class into multiple subclasses of similar sizes as the tail classes and enforce representations to capture the two-layer class hierarchy between the original classes and their subclasses. Since the clustering is conducted in the representation space and updated during the course of training, the subclass labels preserve the semantic substructures of head classes. Meanwhile, it does not overemphasize tail class samples, so each individual instance contribute to the representation learning equally. Hence, our method achieves both the instance- and subclass-balance, while the original class labels are also learned through contrastive learning among subclasses from different classes. We evaluate SBCL over a list of long-tailed benchmark datasets and it achieves the state-of-the-art performance. In addition, we present extensive analyses and ablation studies of SBCL to verify its advantages.

Graph neural networks (GNNs) have pioneered advancements in graph representation learning, exhibiting superior feature learning and performance over multilayer perceptrons (MLPs) when handling graph inputs. However, understanding the feature learning aspect of GNNs is still in its initial stage. This study aims to bridge this gap by investigating the role of graph convolution within the context of feature learning theory in neural networks using gradient descent training. We provide a distinct characterization of signal learning and noise memorization in two-layer graph convolutional networks (GCNs), contrasting them with two-layer convolutional neural networks (CNNs). Our findings reveal that graph convolution significantly augments the benign overfitting regime over the counterpart CNNs, where signal learning surpasses noise memorization, by approximately factor $\sqrt{D}^{q-2}$, with $D$ denoting a node's expected degree and $q$ being the power of the ReLU activation function where $q > 2$. These findings highlight a substantial discrepancy between GNNs and MLPs in terms of feature learning and generalization capacity after gradient descent training, a conclusion further substantiated by our empirical simulations., Comment: 33 pages, 7 figures. We have provided a clearer roadmap

Ranking model plays an essential role in e-commerce search and recommendation. An effective ranking model should give a personalized ranking list for each user according to the user preference. Existing algorithms usually extract a user representation vector from the user behavior sequence, then feed the vector into a feed-forward network (FFN) together with other features for feature interactions, and finally produce a personalized ranking score. Despite tremendous progress in the past, there is still room for improvement. Firstly, the personalized patterns of feature interactions for different users are not explicitly modeled. Secondly, most of existing algorithms have poor personalized ranking results for long-tail users with few historical behaviors due to the data sparsity. To overcome the two challenges, we propose Attention Weighted Mixture of Experts (AW-MoE) with contrastive learning for personalized ranking. Firstly, AW-MoE leverages the MoE framework to capture personalized feature interactions for different users. To model the user preference, the user behavior sequence is simultaneously fed into expert networks and the gate network. Within the gate network, one gate unit and one activation unit are designed to adaptively learn the fine-grained activation vector for experts using an attention mechanism. Secondly, a random masking strategy is applied to the user behavior sequence to simulate long-tail users, and an auxiliary contrastive loss is imposed to the output of the gate network to improve the model generalization for these users. This is validated by a higher performance gain on the long-tail user test set. Experiment results on a JD real production dataset and a public dataset demonstrate the effectiveness of AW-MoE, which significantly outperforms state-of-art methods. Notably, AW-MoE has been successfully deployed in the JD e-commerce search engine, ..., Comment: Accepted by ICDE2023

Medical image data are often limited due to the expensive acquisition and annotation process. Hence, training a deep-learning model with only raw data can easily lead to overfitting. One solution to this problem is to augment the raw data with various transformations, improving the model's ability to generalize to new data. However, manually configuring a generic augmentation combination and parameters for different datasets is non-trivial due to inconsistent acquisition approaches and data distributions. Therefore, automatic data augmentation is proposed to learn favorable augmentation strategies for different datasets while incurring large GPU overhead. To this end, we present a novel method, called Dynamic Data Augmentation (DDAug), which is efficient and has negligible computation cost. Our DDAug develops a hierarchical tree structure to represent various augmentations and utilizes an efficient Monte-Carlo tree searching algorithm to update, prune, and sample the tree. As a result, the augmentation pipeline can be optimized for each dataset automatically. Experiments on multiple Prostate MRI datasets show that our method outperforms the current state-of-the-art data augmentation strategies.

Contrastive Language-Image Pre-training (CLIP) has become the standard for learning cross-modal representations between images and text. Efforts to improve its capabilities typically demand the collection of additional data and retraining with new loss functions. While effective, the added requirements limit their practical use due to the increased resource and time investments needed. In this work, we present HELIP, a cost-effective strategy tailored to enhance the performance of existing CLIP models without the need for training a model from scratch or collecting additional data. Our method allows for effortless integration with existing models' training pipelines, providing an instant boost by training them with selected challenging text-image pairs from their original training datasets. HELIP treats each text-image pair as a single point in the joint vision-language space, identifying those in close proximity as hard pairs. By incorporating the challenging data, pre-trained CLIP models are refined using both the traditional contrastive loss and the newly introduced hard negative margin loss, ensuring the challenging data is fully utilized. On comprehensive benchmarks, HELIP consistently boosts existing models to achieve leading performance. In particular, it improves the zero-shot classification accuracy on ImageNet for SLIP models pre-trained on CC3M, CC12M and YFCC15M datasets. The improvements are 3.05%, 4.47%, and 10.1% respectively, achieved within two epochs of training. In addition, across fine-grained classification datasets, HELIP improves the zero-shot performance of pre-trained CLIP and SLIP by an average of 8.4% and 18.6%, and their linear probe performance by an average of 9.5% and 3.0%., Comment: The code is publicly available at https://github.com/haonan3/HELIP

We propose a stochastic volatility model for time series of curves. It is motivated by dynamics of intraday price curves that exhibit both between days dependence and intraday price evolution. The curves are suitably normalized to stationary in a function space and are functional analogs of point-to-point daily returns. The between curves dependence is modeled by a latent autoregression. The within curves behavior is modeled by a diffusion process. We establish the properties of the model and propose several approaches to its estimation. These approaches are justified by asymptotic arguments that involve an interplay between between the latent autoregression and the intraday diffusions. The asymptotic framework combines the increasing number of daily curves and the refinement of the discrete grid on which each daily curve is observed. Consistency rates for the estimators of the intraday volatility curves are derived as well as the asymptotic normality of the estimators of the latent autoregression. The estimation approaches are further explored and compared by an application to intraday price curves of over seven thousand U.S. stocks and an informative simulation study.

Topological mosaic pattern (TMP) can be formed in two-dimensional (2D) moir\'e superlattices, a set of periodic and spatially separated domains with distinct topologies give rise to periodic edge states on the domain walls. In this study, we demonstrate that these periodic edge states play a crucial role in determining global topological properties. By developing a continuum model for periodic edge states with C6z and C3z rotational symmetry, we predict that a global topological phase transition at the charge neutrality point (CNP) can be driven by the size of domain walls and moir\'e periodicity. The Wannier representation analysis reveals that these periodic edge states are fundamentally chiral px +- ipy orbitals. The interplay between on-site chiral orbital rotation and neighboring hopping among chiral orbitals leads to band inversion and a topological phase transition. Our work establishes a general model for tuning local and global topological phases, paving the way for future research on strongly correlated topological flat minibands within topological mosaic patterns

Effectively extracting inter-frame motion and appearance information is important for video frame interpolation (VFI). Previous works either extract both types of information in a mixed way or elaborate separate modules for each type of information, which lead to representation ambiguity and low efficiency. In this paper, we propose a novel module to explicitly extract motion and appearance information via a unifying operation. Specifically, we rethink the information process in inter-frame attention and reuse its attention map for both appearance feature enhancement and motion information extraction. Furthermore, for efficient VFI, our proposed module could be seamlessly integrated into a hybrid CNN and Transformer architecture. This hybrid pipeline can alleviate the computational complexity of inter-frame attention as well as preserve detailed low-level structure information. Experimental results demonstrate that, for both fixed- and arbitrary-timestep interpolation, our method achieves state-of-the-art performance on various datasets. Meanwhile, our approach enjoys a lighter computation overhead over models with close performance. The source code and models are available at https://github.com/MCG-NJU/EMA-VFI., Comment: Accepted by CVPR 2023

Electrically tuning long-range magnetic orders has been realized in two-dimensional (2D) semiconductors via electrostatic doping. On the other hand, the observations are highly diverse: the transition can be realized by either electrons or holes or both depending on specific materials. Moreover, doped carriers seem to always favor the ferromagnetic (FM) ground state. The mechanism behind those diverse observations remains uncovered. Combining first-principles simulations, we analyze the spin superexchange paths of the correlated d/f orbitals around band edges and assign 2D magnetic semiconductors into three types by their projected density of states (PDOS). We find that each type of PDOS corresponds to a specific carrier-driven magnetic phase transition and the critical doping density and type of carriers can be quantitatively obtained by calculating the superexchange coupling strength. The model results are in good agreements with first-principles calculations and available measurements. After understanding the mechanism, we can design heterostructures to realize the FM to antiferromagnetic transition, which has not been realized before. This model is helpful to understand diverse measurements and expand the degrees of freedom to control long-range magnetic orders in 2D semiconductors., Comment: 20 pages and 5 figures

The security of artificial intelligence (AI) is an important research area towards safe, reliable, and trustworthy AI systems. To accelerate the research on AI security, the Artificial Intelligence Security Competition (AISC) was organized by the Zhongguancun Laboratory, China Industrial Control Systems Cyber Emergency Response Team, Institute for Artificial Intelligence, Tsinghua University, and RealAI as part of the Zhongguancun International Frontier Technology Innovation Competition (https://www.zgc-aisc.com/en). The competition consists of three tracks, including Deepfake Security Competition, Autonomous Driving Security Competition, and Face Recognition Security Competition. This report will introduce the competition rules of these three tracks and the solutions of top-ranking teams in each track., Comment: Technical report of AISC

Existing graph contrastive learning (GCL) techniques typically require two forward passes for a single instance to construct the contrastive loss, which is effective for capturing the low-frequency signals of node features. Such a dual-pass design has shown empirical success on homophilic graphs, but its effectiveness on heterophilic graphs, where directly connected nodes typically have different labels, is unknown. In addition, existing GCL approaches fail to provide strong performance guarantees. Coupled with the unpredictability of GCL approaches on heterophilic graphs, their applicability in real-world contexts is limited. Then, a natural question arises: Can we design a GCL method that works for both homophilic and heterophilic graphs with a performance guarantee? To answer this question, we theoretically study the concentration property of features obtained by neighborhood aggregation on homophilic and heterophilic graphs, introduce the single-pass augmentation-free graph contrastive learning loss based on the property, and provide performance guarantees for the minimizer of the loss on downstream tasks. As a direct consequence of our analysis, we implement the Single-Pass Graph Contrastive Learning method (SP-GCL). Empirically, on 14 benchmark datasets with varying degrees of homophily, the features learned by the SP-GCL can match or outperform existing strong baselines with significantly less computational overhead, which demonstrates the usefulness of our findings in real-world cases., Comment: This article has been accepted for publication by the Transactions on Machine Learning Research. OpenReview at: https://openreview.net/forum?id=244KePn09i

Pipeline parallelism has achieved great success in deploying large-scale transformer models in cloud environments, but has received less attention in edge environments. Unlike in cloud scenarios with high-speed and stable network interconnects, dynamic bandwidth in edge systems can degrade distributed pipeline performance. We address this issue with QuantPipe, a communication-efficient distributed edge system that introduces post-training quantization (PTQ) to compress the communicated tensors. QuantPipe uses adaptive PTQ to change bitwidths in response to bandwidth dynamics, maintaining transformer pipeline performance while incurring limited inference accuracy loss. We further improve the accuracy with a directed-search analytical clipping for integer quantization method (DS-ACIQ), which bridges the gap between estimated and real data distributions. Experimental results show that QuantPipe adapts to dynamic bandwidth to maintain pipeline performance while achieving a practical model accuracy using a wide range of quantization bitwidths, e.g., improving accuracy under 2-bit quantization by 15.85\% on ImageNet compared to naive quantization.

Pre-trained BERT models have achieved impressive accuracy on natural language processing (NLP) tasks. However, their excessive amount of parameters hinders them from efficient deployment on edge devices. Binarization of the BERT models can significantly alleviate this issue but comes with a severe accuracy drop compared with their full-precision counterparts. In this paper, we propose an efficient and robust binary ensemble BERT (BEBERT) to bridge the accuracy gap. To the best of our knowledge, this is the first work employing ensemble techniques on binary BERTs, yielding BEBERT, which achieves superior accuracy while retaining computational efficiency. Furthermore, we remove the knowledge distillation procedures during ensemble to speed up the training process without compromising accuracy. Experimental results on the GLUE benchmark show that the proposed BEBERT significantly outperforms the existing binary BERT models in accuracy and robustness with a 2x speedup on training time. Moreover, our BEBERT has only a negligible accuracy loss of 0.3% compared to the full-precision baseline while saving 15x and 13x in FLOPs and model size, respectively. In addition, BEBERT also outperforms other compressed BERTs in accuracy by up to 6.7%., Comment: To appear in 2023 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2023)

Current state-of-the-art document retrieval solutions mainly follow an index-retrieve paradigm, where the index is hard to be directly optimized for the final retrieval target. In this paper, we aim to show that an end-to-end deep neural network unifying training and indexing stages can significantly improve the recall performance of traditional methods. To this end, we propose Neural Corpus Indexer (NCI), a sequence-to-sequence network that generates relevant document identifiers directly for a designated query. To optimize the recall performance of NCI, we invent a prefix-aware weight-adaptive decoder architecture, and leverage tailored techniques including query generation, semantic document identifiers, and consistency-based regularization. Empirical studies demonstrated the superiority of NCI on two commonly used academic benchmarks, achieving +21.4% and +16.8% relative enhancement for Recall@1 on NQ320k dataset and R-Precision on TriviaQA dataset, respectively, compared to the best baseline method., Comment: 19 pages, 6 figures, accepted by NeurIPS 2022

Programmatic Weak Supervision (PWS) aggregates the source votes of multiple weak supervision sources into probabilistic training labels, which are in turn used to train an end model. With its increasing popularity, it is critical to have some tool for users to understand the influence of each component (e.g., the source vote or training data) in the pipeline and interpret the end model behavior. To achieve this, we build on Influence Function (IF) and propose source-aware IF, which leverages the generation process of the probabilistic labels to decompose the end model's training objective and then calculate the influence associated with each (data, source, class) tuple. These primitive influence score can then be used to estimate the influence of individual component of PWS, such as source vote, supervision source, and training data. On datasets of diverse domains, we demonstrate multiple use cases: (1) interpreting incorrect predictions from multiple angles that reveals insights for debugging the PWS pipeline, (2) identifying mislabeling of sources with a gain of 9%-37% over baselines, and (3) improving the end model's generalization performance by removing harmful components in the training objective (13%-24% better than ordinary IF)., Comment: 21 pages

Graph contrastive learning (GCL) is the most representative and prevalent self-supervised learning approach for graph-structured data. Despite its remarkable success, existing GCL methods highly rely on an augmentation scheme to learn the representations invariant across different augmentation views. In this work, we revisit such a convention in GCL through examining the effect of augmentation techniques on graph data via the lens of spectral theory. We found that graph augmentations preserve the low-frequency components and perturb the middle-and high-frequency components of the graph, which contributes to the success of GCL algorithms on homophilic graphs but hinder its application on heterophilic graphs, due to the high-frequency preference of heterophilic data. Motivated by this, we propose a novel, theoretically-principled, and augmentation-free GCL method, named AF-GCL, that (1) leverages the features aggregated by Graph Neural Network to construct the self-supervision signal instead of augmentations and therefore (2) is less sensitive to the graph homophily degree. Theoretically, We present the performance guarantee for AF-GCL as well as an analysis for understanding the efficacy of AF-GCL. Extensive experiments on 14 benchmark datasets with varying degrees of heterophily show that AF-GCL presents competitive or better performance on homophilic graphs and outperforms all existing state-of-the-art GCL methods on heterophilic graphs with significantly less computational overhead.

In this paper, we focus on the pattern reconfigurable multiple-input multiple-output (PR-MIMO), a technique that has the potential to bridge the gap between electro-magnetics and communications towards the emerging Electro-magnetic Information Theory (EIT). Specifically, we focus on the pattern design problem aimed at maximizing the channel capacity for reconfigurable MIMO communication systems, where we firstly introduce the matrix representation of PR-MIMO and further formulate a pattern design problem. We decompose the pattern design into two steps, i.e., the correlation modification process to optimize the correlation structure of the channel, followed by the power allocation process to improve the channel quality based on the optimized channel structure. For the correlation modification process, we propose a sequential optimization framework with eigenvalue decomposition to obtain near-optimal solutions. For the power allocation process, we provide a closed-form power allocation scheme to redistribute the transmission power among the modified subchannels. Numerical results show that the proposed pattern design scheme offers significant improvements over legacy MIMO systems, which motivates the application of PR-MIMO in wireless communication systems., Comment: This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible

Reconfigurable multiple-input multiple-output can provide performance gains over traditional MIMO by reshaping the channels, i.e., introducing more channel realizations. In this paper, we focus on the achievable rate maximization pattern design for reconfigurable MIMO systems. Firstly, we introduce the matrix representation of pattern reconfigurable MIMO (PR-MIMO), based on which a pattern design problem is formulated. To further reveal the effect of the radiation pattern on the wireless channel, we consider pattern design for both the single-pattern case where the optimized radiation pattern is the same for all the antenna elements, and the multi-pattern case where different antenna elements can adopt different radiation patterns. For the single-pattern case, we show that the pattern design is equivalent to a redistribution of gains among all scattering paths, and an eigenvalue optimization based solution is obtained. For the multi-pattern case, we propose a sequential optimization framework with manifold optimization and eigenvalue decomposition to obtain near-optimal solutions. Numerical results validate the superiority of PR-MIMO systems over traditional MIMO in terms of achievable rate, and also show the effectiveness of the proposed solutions., Comment: This work has been accepted by IEEE Transactions on Wireless Communications

Monolayer graphene embedded with transition metal nitride (i.e., $X$N$_4$) has been experimentally synthesized recently, where a transition metal atom together with four nitrogen atoms as a unit are embedded in graphene to form a stable planar single-atom-thick structure. We provide a systematic study on the structural, electronic and topological properties of these $X$N$_4$-embedded graphene by utilizing both first-principles calculations and tight-binding model. We find that $X$N$_4$-embedded graphene ($X$=Pt, Ir, Rh, Os) can open topologically nontrivial band gaps that host \emph{two-dimensional} $\mathbb{Z}_2$ topological insulators. We further show that the low-energy bands near the band gaps can be perfectly captured by a modified Kane-Mele model Hamiltonian. Our work not only provides concrete two-dimensional materials that are very rare to realize \emph{two-dimensional} $\mathbb{Z}_2$ topological insulators, but also makes the graphene system to be realistic in hosting Kane-Mele type $\mathbb{Z}_2$ topological insulators.