arXiv subjects

Shan Yu

Publications and source records attributed to Shan Yu.

At least 55 records · Page 3Linked to original sources

arXiv2024

Spatial Heterogeneous Additive Partial Linear Model: A Joint Approach of Bivariate Spline and Forest Lasso

Identifying spatial heterogeneous patterns has attracted a surge of research interest in recent years, due to its important applications in various scientific and engineering fields. In practice the spatially heterogeneous components are often mixed with components which are spatially smooth, making the task of identifying the heterogeneous regions more challenging. In this paper, we develop an efficient clustering approach to identify the model heterogeneity of the spatial additive partial linear model. Specifically, we aim to detect the spatially contiguous clusters based on the regression coefficients while introducing a spatially varying intercept to deal with the smooth spatial effect. On the one hand, to approximate the spatial varying intercept, we use the method of bivariate spline over triangulation, which can effectively handle the data from a complex domain. On the other hand, a novel fusion penalty termed the forest lasso is proposed to reveal the spatial clustering pattern. Our proposed fusion penalty has advantages in both the estimation and computation efficiencies when dealing with large spatial data. Theoretically properties of our estimator are established, and simulation results show that our approach can achieve more accurate estimation with a limited computation cost compared with the existing approaches. To illustrate its practical use, we apply our approach to analyze the spatial pattern of the relationship between land surface temperature measured by satellites and air temperature measured by ground stations in the United States.

stat.ME↗

arXiv2024

Nonparametric Automatic Differentiation Variational Inference with Spline Approximation

Automatic Differentiation Variational Inference (ADVI) is efficient in learning probabilistic models. Classic ADVI relies on the parametric approach to approximate the posterior. In this paper, we develop a spline-based nonparametric approximation approach that enables flexible posterior approximation for distributions with complicated structures, such as skewness, multimodality, and bounded support. Compared with widely-used nonparametric variational inference methods, the proposed method is easy to implement and adaptive to various data structures. By adopting the spline approximation, we derive a lower bound of the importance weighted autoencoder and establish the asymptotic consistency. Experiments demonstrate the efficiency of the proposed method in approximating complex posterior distributions and improving the performance of generative models with incomplete data.

stat.ML↗

arXiv2024

TSSS: A Novel Triangulated Spherical Spline Smoothing for Surface-based Data

Surface-based data is commonly observed in diverse practical applications spanning various fields. In this paper, we introduce a novel nonparametric method to discover the underlying signals from data distributed on complex surface-based domains. Our approach involves a penalized spline estimator defined on a triangulation of surface patches, which enables effective signal extraction and recovery. The proposed method offers several advantages over existing methods, including superior handling of "leakage" or "boundary effects" over complex domains, enhanced computational efficiency, and potential applications in analyzing sparse and irregularly distributed data on complex objects. We provide rigorous theoretical guarantees for the proposed method, including convergence rates of the estimator in both the $L_2$ and supremum norms, as well as the asymptotic normality of the estimator. We also demonstrate that the convergence rates achieved by our estimation method are optimal within the framework of nonparametric estimation. Furthermore, we introduce a bootstrap method to quantify the uncertainty associated with the proposed estimators accurately. The superior performance of the proposed method is demonstrated through simulation experiments and data applications on cortical surface functional magnetic resonance imaging data and oceanic near-surface atmospheric data.

stat.ME↗

arXiv2023

AnyOKP: One-Shot and Instance-Aware Object Keypoint Extraction with Pretrained ViT

Towards flexible object-centric visual perception, we propose a one-shot instance-aware object keypoint (OKP) extraction approach, AnyOKP, which leverages the powerful representation ability of pretrained vision transformer (ViT), and can obtain keypoints on multiple object instances of arbitrary category after learning from a support image. An off-the-shelf petrained ViT is directly deployed for generalizable and transferable feature extraction, which is followed by training-free feature enhancement. The best-prototype pairs (BPPs) are searched for in support and query images based on appearance similarity, to yield instance-unaware candidate keypoints.Then, the entire graph with all candidate keypoints as vertices are divided to sub-graphs according to the feature distributions on the graph edges. Finally, each sub-graph represents an object instance. AnyOKP is evaluated on real object images collected with the cameras of a robot arm, a mobile robot, and a surgical robot, which not only demonstrates the cross-category flexibility and instance awareness, but also show remarkable robustness to domain shift and viewpoint change.

cs.CV↗

arXiv2023

Ethosight: A Reasoning-Guided Iterative Learning System for Nuanced Perception based on Joint-Embedding & Contextual Label Affinity

Traditional computer vision models often necessitate extensive data acquisition, annotation, and validation. These models frequently struggle in real-world applications, resulting in high false positive and negative rates, and exhibit poor adaptability to new scenarios, often requiring costly retraining. To address these issues, we present Ethosight, a flexible and adaptable zero-shot video analytics system. Ethosight begins from a clean slate based on user-defined video analytics, specified through natural language or keywords, and leverages joint embedding models and reasoning mechanisms informed by ontologies such as WordNet and ConceptNet. Ethosight operates effectively on low-cost edge devices and supports enhanced runtime adaptation, thereby offering a new approach to continuous learning without catastrophic forgetting. We provide empirical validation of Ethosight's promising effectiveness across diverse and complex use cases, while highlighting areas for further improvement. A significant contribution of this work is the release of all source code and datasets to enable full reproducibility and to foster further innovation in both the research and commercial domains.

cs.CV↗

arXiv2023

Emergence of Symbols in Neural Networks for Semantic Understanding and Communication

The capacity to generate meaningful symbols and effectively employ them for advanced cognitive processes, such as communication, reasoning, and planning, constitutes a fundamental and distinctive aspect of human intelligence. Existing deep neural networks still notably lag human capabilities in terms of generating symbols for higher cognitive functions. Here, we propose a solution (symbol emergence artificial network (SEA-net)) to endow neural networks with the ability to create symbols, understand semantics, and achieve communication. SEA-net generates symbols that dynamically configure the network to perform specific tasks. These symbols capture compositional semantic information that allows the system to acquire new functions purely by symbolic manipulation or communication. In addition, these self-generated symbols exhibit an intrinsic structure resembling that of natural language, suggesting a common framework underlying the generation and understanding of symbols in both human brains and artificial neural networks. We believe that the proposed framework will be instrumental in producing more capable systems that can synergize the strengths of connectionist and symbolic approaches for artificial intelligence (AI).

cs.AI↗

arXiv2023

Adaptive Data Augmentation for Contrastive Learning

In computer vision, contrastive learning is the most advanced unsupervised learning framework. Yet most previous methods simply apply fixed composition of data augmentations to improve data efficiency, which ignores the changes in their optimal settings over training. Thus, the pre-determined parameters of augmentation operations cannot always fit well with an evolving network during the whole training period, which degrades the quality of the learned representations. In this work, we propose AdDA, which implements a closed-loop feedback structure to a generic contrastive learning network. AdDA works by allowing the network to adaptively adjust the augmentation compositions according to the real-time feedback. This online adjustment helps maintain the dynamic optimal composition and enables the network to acquire more generalizable representations with minimal computational overhead. AdDA achieves competitive results under the common linear protocol on ImageNet-100 classification (+1.11% on MoCo v2).

cs.CV↗

arXiv2023

Per-RMAP: Feasibility-Seeking and Superiorization Methods for Floorplanning with I/O Assignment

The feasibility-seeking approach provides a systematic scheme to manage and solve complex constraints for continuous problems, and we explore it for the floorplanning problems with increasingly heterogeneous constraints. The classic legality constraints can be formulated as the union of convex sets. However, the convergence of conventional projection-based algorithms is not guaranteed as the constrain sets are non-convex. In this work, we propose a resetting strategy to greatly eliminate the the divergence issue of the projection-based algorithm for the feasibility-seeking formulation. Furthermore, the superiorization methodology (SM), which lies between feasibility-seeking and constrained optimization, is firstly applied to floorplanning. The SM uses perturbations to steer the feasibility-seeking algorithm to a feasible solution with shorter total wirelength. The proposed flow is extendable to tackle various constraints and variants of floorplanning problems, e.g., floorplanning with I/O assignment problems. We have evaluated the proposed algorithm on the MCNC benchmarks. We can obtain legal floorplans only two times slower than the branch-and-bound method in its current prototype using MATLAB, with only 3% wirelength inferior to the optimal results. We evaluate the effectiveness of the flow by considering the constraints of I/O assignment, and our algorithm achieve 8% improvement on wirelength.

math.OC↗

arXiv2023

TQ-Net: Mixed Contrastive Representation Learning For Heterogeneous Test Questions

Recently, more and more people study online for the convenience of access to massive learning materials (e.g. test questions/notes), thus accurately understanding learning materials became a crucial issue, which is essential for many educational applications. Previous studies focus on using language models to represent the question data. However, test questions (TQ) are usually heterogeneous and multi-modal, e.g., some of them may only contain text, while others half contain images with information beyond their literal description. In this context, both supervised and unsupervised methods are difficult to learn a fused representation of questions. Meanwhile, this problem cannot be solved by conventional methods such as image caption, as the images may contain information complementary rather than duplicate to the text. In this paper, we first improve previous text-only representation with a two-stage unsupervised instance level contrastive based pre-training method (MCL: Mixture Unsupervised Contrastive Learning). Then, TQ-Net was proposed to fuse the content of images to the representation of heterogeneous data. Finally, supervised contrastive learning was conducted on relevance prediction-related downstream tasks, which helped the model to learn the representation of questions effectively. We conducted extensive experiments on question-based tasks on large-scale, real-world datasets, which demonstrated the effectiveness of TQ-Net and improve the precision of downstream applications (e.g. similar questions +2.02% and knowledge point prediction +7.20%). Our code will be available, and we will open-source a subset of our data to promote the development of relative studies.

cs.CL↗

arXiv2023

Nonparametric Regression for 3D Point Cloud Learning

Over the past two decades, we have seen an exponentially increased amount of point clouds collected with irregular shapes in various areas. Motivated by the importance of solid modeling for point clouds, we develop a novel and efficient smoothing tool based on multivariate splines over the tetrahedral partitions to extract the underlying signal and build up a 3D solid model from the point cloud. The proposed smoothing method can denoise or deblur the point cloud effectively and provide a multi-resolution reconstruction of the actual signal. In addition, it can handle sparse and irregularly distributed point clouds and recover the underlying trajectory. The proposed smoothing and interpolation method also provides a natural way of numerosity data reduction. Furthermore, we establish the theoretical guarantees of the proposed method. Specifically, we derive the convergence rate and asymptotic normality of the proposed estimator and illustrate that the convergence rate achieves the optimal nonparametric convergence rate. Through extensive simulation studies and a real data example, we demonstrate the superiority of the proposed method over traditional smoothing methods in terms of estimation accuracy and efficiency of data reduction.

stat.CO↗

arXiv2023

AI of Brain and Cognitive Sciences: From the Perspective of First Principles

Nowadays, we have witnessed the great success of AI in various applications, including image classification, game playing, protein structure analysis, language translation, and content generation. Despite these powerful applications, there are still many tasks in our daily life that are rather simple to humans but pose great challenges to AI. These include image and language understanding, few-shot learning, abstract concepts, and low-energy cost computing. Thus, learning from the brain is still a promising way that can shed light on the development of next-generation AI. The brain is arguably the only known intelligent machine in the universe, which is the product of evolution for animals surviving in the natural environment. At the behavior level, psychology and cognitive sciences have demonstrated that human and animal brains can execute very intelligent high-level cognitive functions. At the structure level, cognitive and computational neurosciences have unveiled that the brain has extremely complicated but elegant network forms to support its functions. Over years, people are gathering knowledge about the structure and functions of the brain, and this process is accelerating recently along with the initiation of giant brain projects worldwide. Here, we argue that the general principles of brain functions are the most valuable things to inspire the development of AI. These general principles are the standard rules of the brain extracting, representing, manipulating, and retrieving information, and here we call them the first principles of the brain. This paper collects six such first principles. They are attractor network, criticality, random network, sparse coding, relational memory, and perceptual learning. On each topic, we review its biological background, fundamental property, potential application to AI, and future development.

q-bio.NC↗

arXiv2022

BigDL 2.0: Seamless Scaling of AI Pipelines from Laptops to Distributed Cluster

Most AI projects start with a Python notebook running on a single laptop; however, one usually needs to go through a mountain of pains to scale it to handle larger dataset (for both experimentation and production deployment). These usually entail many manual and error-prone steps for the data scientists to fully take advantage of the available hardware resources (e.g., SIMD instructions, multi-processing, quantization, memory allocation optimization, data partitioning, distributed computing, etc.). To address this challenge, we have open sourced BigDL 2.0 at https://github.com/intel-analytics/BigDL/ under Apache 2.0 license (combining the original BigDL and Analytics Zoo projects); using BigDL 2.0, users can simply build conventional Python notebooks on their laptops (with possible AutoML support), which can then be transparently accelerated on a single node (with up-to 9.6x speedup in our experiments), and seamlessly scaled out to a large cluster (across several hundreds servers in real-world use cases). BigDL 2.0 has already been adopted by many real-world users (such as Mastercard, Burger King, Inspur, etc.) in production.

cs.LG↗

arXiv2022

Recursive Least-Squares Estimator-Aided Online Learning for Visual Tracking

Tracking visual objects from a single initial exemplar in the testing phase has been broadly cast as a one-/few-shot problem, i.e., one-shot learning for initial adaptation and few-shot learning for online adaptation. The recent few-shot online adaptation methods incorporate the prior knowledge from large amounts of annotated training data via complex meta-learning optimization in the offline phase. This helps the online deep trackers to achieve fast adaptation and reduce overfitting risk in tracking. In this paper, we propose a simple yet effective recursive least-squares estimator-aided online learning approach for few-shot online adaptation without requiring offline training. It allows an in-built memory retention mechanism for the model to remember the knowledge about the object seen before, and thus the seen data can be safely removed from training. This also bears certain similarities to the emerging continual learning field in preventing catastrophic forgetting. This mechanism enables us to unveil the power of modern online deep trackers without incurring too much extra computational cost. We evaluate our approach based on two networks in the online learning families for tracking, i.e., multi-layer perceptrons in RT-MDNet and convolutional neural networks in DiMP. The consistent improvements on several challenging tracking benchmarks demonstrate its effectiveness and efficiency.

cs.CV↗

arXiv2021

Continual Learning of Context-dependent Processing in Neural Networks

Deep neural networks (DNNs) are powerful tools in learning sophisticated but fixed mapping rules between inputs and outputs, thereby limiting their application in more complex and dynamic situations in which the mapping rules are not kept the same but changing according to different contexts. To lift such limits, we developed a novel approach involving a learning algorithm, called orthogonal weights modification (OWM), with the addition of a context-dependent processing (CDP) module. We demonstrated that with OWM to overcome the problem of catastrophic forgetting, and the CDP module to learn how to reuse a feature representation and a classifier for different contexts, a single network can acquire numerous context-dependent mapping rules in an online and continual manner, with as few as $\sim$10 samples to learn each. This should enable highly compact systems to gradually learn myriad regularities of the real world and eventually behave appropriately within it.

cs.LG↗

arXiv2021

Multivariate Spline Estimation and Inference for Image-On-Scalar Regression

Motivated by recent data analyses in biomedical imaging studies, we consider a class of image-on-scalar regression models for imaging responses and scalar predictors. We propose using flexible multivariate splines over triangulations to handle the irregular domain of the objects of interest on the images, as well as other characteristics of images. The proposed estimators of the coefficient functions are proved to be root-n consistent and asymptotically normal under some regularity conditions. We also provide a consistent and computationally efficient estimator of the covariance function. Asymptotic pointwise confidence intervals and data-driven simultaneous confidence corridors for the coefficient functions are constructed. Our method can simultaneously estimate and make inferences on the coefficient functions while incorporating spatial heterogeneity and spatial correlation. A highly efficient and scalable estimation algorithm is developed. Monte Carlo simulation studies are conducted to examine the finite-sample performance of the proposed method, which is then applied to the spatially normalized positron emission tomography data of the Alzheimer's Disease Neuroimaging Initiative.

stat.ME↗

arXiv2021

Intra-Class Uncertainty Loss Function for Classification

Most classification models can be considered as the process of matching templates. However, when intra-class uncertainty/variability is not considered, especially for datasets containing unbalanced classes, this may lead to classification errors. To address this issue, we propose a loss function with intra-class uncertainty following Gaussian distribution. Specifically, in our framework, the features extracted by deep networks of each class are characterized by independent Gaussian distribution. The parameters of distribution are learned with a likelihood regularization along with other network parameters. The means of the Gaussian play a similar role as the center anchor in existing methods, and the variance describes the uncertainty of different classes. In addition, similar to the inter-class margin in traditional loss functions, we introduce a margin to intra-class uncertainty to make each cluster more compact and reduce the imbalance of feature distribution from different categories. Based on MNIST, CIFAR, ImageNet, and Long-tailed CIFAR analyses, the proposed approach shows improved classification performance, through learning a better class representation.

cs.CV↗

arXiv2021

EEG-Based Brain-Computer Interfaces Are Vulnerable to Backdoor Attacks

Research and development of electroencephalogram (EEG) based brain-computer interfaces (BCIs) have advanced rapidly, partly due to deeper understanding of the brain and wide adoption of sophisticated machine learning approaches for decoding the EEG signals. However, recent studies have shown that machine learning algorithms are vulnerable to adversarial attacks. This article proposes to use narrow period pulse for poisoning attack of EEG-based BCIs, which is implementable in practice and has never been considered before. One can create dangerous backdoors in the machine learning model by injecting poisoning samples into the training set. Test samples with the backdoor key will then be classified into the target class specified by the attacker. What most distinguishes our approach from previous ones is that the backdoor key does not need to be synchronized with the EEG trials, making it very easy to implement. The effectiveness and robustness of the backdoor attack approach is demonstrated, highlighting a critical security concern for EEG-based BCIs and calling for urgent attention to address it.

cs.CR↗

arXiv2020

Spatiotemporal Dynamics, Nowcasting and Forecasting of COVID-19 in the United States

Epidemic modeling is an essential tool to understand the spread of the novel coronavirus and ultimately assist in disease prevention, policymaking, and resource allocation. In this article, we establish a state of the art interface between classic mathematical and statistical models and propose a novel space-time epidemic modeling framework to study the spatial-temporal pattern in the spread of infectious disease. We propose a quasi-likelihood approach via the penalized spline approximation and alternatively reweighted least-squares technique to estimate the model. Furthermore, we provide a short-term and long-term county-level prediction of the infected/death count for the U.S. by accounting for the control measures, health service resources, and other local features. Utilizing spatiotemporal analysis, our proposed model enhances the dynamics of the epidemiological mechanism and dissects the spatiotemporal structure of the spreading disease. To assess the uncertainty associated with the prediction, we develop a projection band based on the envelope of the bootstrap forecast paths. The performance of the proposed method is evaluated by a simulation study. We apply the proposed method to model and forecast the spread of COVID-19 at both county and state levels in the United States.

stat.AP↗