Deyu Meng
YOU?
Author Swipe
View article: Feed Two Birds with One Scone: Exploiting Function-Space Regularization for Both OOD Robustness and ID Fine-Tuning Performance
Feed Two Birds with One Scone: Exploiting Function-Space Regularization for Both OOD Robustness and ID Fine-Tuning Performance Open
Robust fine-tuning aims to achieve competitive in-distribution (ID) performance while maintaining the out-of-distribution (OOD) robustness of a pre-trained model when transferring it to a downstream task. To remedy this, most robust fine-t…
View article: Beyond Low-rankness: Guaranteed Matrix Recovery via Modified Nuclear Norm
Beyond Low-rankness: Guaranteed Matrix Recovery via Modified Nuclear Norm Open
The nuclear norm (NN) has been widely explored in matrix recovery problems, such as Robust PCA and matrix completion, leveraging the inherent global low-rank structure of the data. In this study, we introduce a new modified nuclear norm (M…
View article: Generative Latent Kernel Modeling for Blind Motion Deblurring
Generative Latent Kernel Modeling for Blind Motion Deblurring Open
Deep prior-based approaches have demonstrated remarkable success in blind motion deblurring (BMD) recently. These methods, however, are often limited by the high non-convexity of the underlying optimization process in BMD, which leads to e…
View article: MolUNet++: Adaptive-grained Explicit Substructure and Interaction Aware Molecular Representation Learning
MolUNet++: Adaptive-grained Explicit Substructure and Interaction Aware Molecular Representation Learning Open
Molecular representation learning is a critical task in AI-driven drug development. While graph neural networks (GNNs) have demonstrated strong performance and gained widespread adoption in this field, efficiently extracting and explicitly…
View article: The evolution of high-order genome architecture revealed from 1,000 species
The evolution of high-order genome architecture revealed from 1,000 species Open
SUMMARY Spatial genome organization plays a crucial regulatory role, but its evolution remains unknown. Leveraging Hi-C data from 1,025 species, we trace the evolutionary trajectories of 3D genome, through two higher-order architectures, ‘…
View article: Flexible Dual‐Modal Sensors Based on Single‐Crystalline Silicon Membranes for Continuous Monitoring of Photoplethysmography and Skin Temperature
Flexible Dual‐Modal Sensors Based on Single‐Crystalline Silicon Membranes for Continuous Monitoring of Photoplethysmography and Skin Temperature Open
Flexible dual‐modal sensors that can monitor two signals play important roles in biomedical applications, along with the recent progresses of epidermal or bioimplantable electronic devices with diagnostic or therapeutic functionalities. Ho…
View article: RS-MTDF: Multi-Teacher Distillation and Fusion for Remote Sensing Semi-Supervised Semantic Segmentation
RS-MTDF: Multi-Teacher Distillation and Fusion for Remote Sensing Semi-Supervised Semantic Segmentation Open
Semantic segmentation in remote sensing images is crucial for various applications, yet its performance is heavily reliant on large-scale, high-quality pixel-wise annotations, which are notoriously expensive and time-consuming to acquire. …
View article: Improving Memory Efficiency for Training KANs via Meta Learning
Improving Memory Efficiency for Training KANs via Meta Learning Open
Inspired by the Kolmogorov-Arnold representation theorem, KANs offer a novel framework for function approximation by replacing traditional neural network weights with learnable univariate functions. This design demonstrates significant pot…
View article: Data-Distill-Net: A Data Distillation Approach Tailored for Reply-based Continual Learning
Data-Distill-Net: A Data Distillation Approach Tailored for Reply-based Continual Learning Open
Replay-based continual learning (CL) methods assume that models trained on a small subset can also effectively minimize the empirical risk of the complete dataset. These methods maintain a memory buffer that stores a sampled subset of data…
View article: Symmetry Breaking in Neural Network Optimization: Insights from Input Dimension Expansion
Symmetry Breaking in Neural Network Optimization: Insights from Input Dimension Expansion Open
Understanding how neural networks learn and optimize remains a central point in machine learning, with implications for designing better models. While techniques like dropout and batch normalization are widely used, the underlying principl…
View article: SegEarth-R1: Geospatial Pixel Reasoning via Large Language Model
SegEarth-R1: Geospatial Pixel Reasoning via Large Language Model Open
Remote sensing has become critical for understanding environmental dynamics, urban planning, and disaster management. However, traditional remote sensing workflows often rely on explicit segmentation or detection methods, which struggle to…
View article: SpaKnit: correlation subspace learning for integrating spatial multi-omics data
SpaKnit: correlation subspace learning for integrating spatial multi-omics data Open
Integrating spatial multi-omics data presents significant challenges, particularly in uncovering the spatial patterns of cells and deciphering the real regulatory mechanisms among various omics. These insights are critical for harnessing t…
View article: Deep Rank-One Tensor Functional Factorization for Multi-Dimensional Data Recovery
Deep Rank-One Tensor Functional Factorization for Multi-Dimensional Data Recovery Open
Many real-world data are inherently multi-dimensional, e.g., color images, videos, and hyperspectral images. How to effectively and compactly represent these multi-dimensional data within a unified framework is an important pursuit. Previo…
View article: Sequential Monte Carlo with Gaussian Mixture Approximation for Infinite-Dimensional Statistical Inverse Problems
Sequential Monte Carlo with Gaussian Mixture Approximation for Infinite-Dimensional Statistical Inverse Problems Open
By formulating the inverse problem of partial differential equations (PDEs) as a statistical inference problem, the Bayesian approach provides a general framework for quantifying uncertainties. In the inverse problem of PDEs, parameters ar…
View article: On the uncertainty principle of neural networks
On the uncertainty principle of neural networks Open
In this study, we explore the inherent trade-off between accuracy and robustness in neural networks, drawing an analogy to the uncertainty principle in quantum mechanics. We propose that neural networks are subject to an uncertainty relati…
View article: Exploring Imbalanced Annotations for Effective In-Context Learning
Exploring Imbalanced Annotations for Effective In-Context Learning Open
Large language models (LLMs) have shown impressive performance on downstream tasks through in-context learning (ICL), which heavily relies on the demonstrations selected from annotated datasets. However, these datasets often exhibit long-t…
View article: DynamicEarth: How Far are We from Open-Vocabulary Change Detection?
DynamicEarth: How Far are We from Open-Vocabulary Change Detection? Open
Monitoring Earth's evolving land covers requires methods capable of detecting changes across a wide range of categories and contexts. Existing change detection methods are hindered by their dependency on predefined classes, reducing their …
View article: IV-tuning: Parameter-Efficient Transfer Learning for Infrared-Visible Tasks
IV-tuning: Parameter-Efficient Transfer Learning for Infrared-Visible Tasks Open
Existing infrared and visible (IR-VIS) methods inherit the general representations of Pre-trained Visual Models (PVMs) to facilitate complementary learning. However, our analysis indicates that under the full fine-tuning paradigm, the feat…
View article: Is AI Robust Enough for Scientific Research?
Is AI Robust Enough for Scientific Research? Open
We uncover a phenomenon largely overlooked by the scientific community utilizing AI: neural networks exhibit high susceptibility to minute perturbations, resulting in significant deviations in their outputs. Through an analysis of five div…
View article: AeroGen: Enhancing Remote Sensing Object Detection with Diffusion-Driven Data Generation
AeroGen: Enhancing Remote Sensing Object Detection with Diffusion-Driven Data Generation Open
Remote sensing image object detection (RSIOD) aims to identify and locate specific objects within satellite or aerial imagery. However, there is a scarcity of labeled data in current RSIOD datasets, which significantly limits the performan…
View article: Graph Domain Adaptation with Dual-branch Encoder and Two-level Alignment for Whole Slide Image-based Survival Prediction
Graph Domain Adaptation with Dual-branch Encoder and Two-level Alignment for Whole Slide Image-based Survival Prediction Open
In recent years, histopathological whole slide image (WSI)- based survival analysis has attracted much attention in medical image analysis. In practice, WSIs usually come from different hospitals or laboratories, which can be seen as diffe…