Large language models (LLMs) have emerged as powerful tools for biological sequence analysis. However, their applicability to the transcriptional regulation of long non-coding RNAs (IncRNAs) remains underexplored due to the complexity and diversity of IncRNA sequences, combined with limited knowledge of their regulatory mechanisms and functional characteristics. In this study, we systematically evaluate both singletask and multi-task fine-tuning strategies of genome foundation models across four tasks designed to capture increasing biological complexity. By fine-tuning genome foundation models on a series of progressively complex tasks, each designed to closely mimic the complexities of IncRNA classification, we explore how task complexity impacts model performance and biological interpretability. Our findings reveal that while foundation models capture promoter-specific signals, task complexity, input length, and data quality significantly influence performance. Notably, the most challenging task, promoter sequence classification of IncRNA from coding genes, remains limited by primary sequence features alone. Through performance evaluation and attentionbased interpretability analysis, this work provides key insights into the potential and limitations of current LLM-based models and offers a foundation for future development of integrative models in IncRNA regulation analysis. The dataset is available at https://github.com/wangwe90/IncRNA-promoter-sequence-detection-dataset

This work tackles an intriguing and fundamental open challenge in representation learning: Given a well-trained deep learning model, can it be reprogrammed to enhance its robustness against adversarial or noisy input perturbations without altering its parameters? To explore this, we revisit the core feature transformation mechanism in representation learning and propose a novel non-linear robust pattern matching technique as a robust alternative. Furthermore, we introduce three model reprogramming paradigms to offer flexible control of robustness under different efficiency requirements. Comprehensive experiments and ablation studies across diverse learning models ranging from basic linear model and MLPs to shallow and modern deep ConvNets demonstrate the effectiveness of our approaches. This work not only opens a promising and orthogonal direction for improving adversarial defenses in deep learning beyond existing methods but also provides new insights into designing more resilient AI systems with robust statistics. Our implementation is available at https://github.com/chris-hzc/Robustness-Reprogramming.

Diffusion models have emerged as powerful generative models, but their high computational cost in iterative sampling remains a significant bottleneck. In this work, we present an in-depth and insightful study of state-of-the-art acceleration techniques for diffusion models, including caching and quantization, and reveal their limitations in computation error and generation quality. To break these limits, this work introduces Modulated Diffusion (MoDiff), an innovative, rigorous, and principled framework that accelerates generative modeling through modulated quantization and error compensation. MoDiff not only inherits the advantages of existing caching and quantization methods but also serves as a general framework to accelerate all diffusion models. The advantages of MoDiff are supported by solid theoretical insight and analysis. In addition, extensive experiments on CIFAR-10 and LSUN demonstrate that MoDiff significantly reduces activation quantization from 8 bits to 3 bits without performance degradation in post-training quantization (PTQ). Our code implementation is available at https://github.com/WeizhiGao/MoDiff.

In-wheel motor drive (IWMD) presents a pivotal solution to surmount obstacles in the electrification of transportation technology. However, the increased unsprung mass due to IWMD deteriorates ride comfort. Various control strategies have been proposed to address this issue. Road irregularities limit control systems, with few adapting to changing conditions. Most rely on vision-based monitoring, offering limited road profile data without past or future predictions to improve performance. This study presents an adaptive fractional-order proportional integral derivative (FOPID) controller for a semi-active suspension system of a vehicle using IWMD. The proposed control strategy leverages cloud computing, storing and processing road preview data in a cloud database. Adaptive particle swarm optimization (APSO) optimizes control parameters for enhanced performance, with real-time optimization enabling adaptive acquisition of optimal parameters. Body acceleration (BA), suspension working space (SWS), and dynamic tire load (DTL) are used to assess suspension performance. Simulation results confirm the proposed method's effectiveness in improving dynamic performance.

The rapid development of IoT technology has promoted the integration and networked fusion of massive heterogeneous sensors. However, traditional Gaussian-Markov frameworks struggle to characterize nonlinear long-term correlated errors within systems and network cooperative effects, limiting inter-node interoperability. To address this, we propose a generalized nonlinear stochastic process (GNSP) theoretical framework, constructing nonlinear operators with long-term memory characteristics, and employing fractional calculus tools to achieve unified analysis of multiple error types within and across nodes. Based on this, we derive distributed error propagation laws on Lie group spaces, establish boundary criteria based on Wasserstein metrics, and improve nonlinear filters to design a dual-layer fusion architecture combining single-node hybrid adaptive filtering and inter-node collaborative optimization. Using vehicle networking GNSS/INS systems as an example, experimental results show that the proposed single-node strategy improves position, velocity, and attitude accuracy by an average of 11.71%, 9.82%, and 12.83% respectively in complex environments such as urban canyons and mountainous forest areas. Meanwhile, through inter-node parameter sharing and cooperative strategies of federated filtering, further V2V collaboration improves performance by 8.5%-12.7%, maintaining average improvements of 16.85%, 18.37%, and 21.14% even under limited communication conditions. This effectively addresses the challenges of distributed sensing errors in IoT environments.

Recently, polynomial graph filter learning (PGFL) has demonstrated promising performance for modeling graph signals in Graph Neural Networks (GNNs) on both homophilic and heterophilic graphs, owning to their flexibility and expressiveness. In this work, we conduct a novel preliminary study to explore the potential and limitations of PGFL, revealing a severe overfitting issue. To improve the effectiveness of PGFL, we propose Auto-Polynomial, a novel and general automated PGFL framework that efficiently learns better filters capable of adapting to various complex graph signals. Comprehensive experiments and ablation studies demonstrate significant and consistent performance improvements on both homophilic and heterophilic graphs across multiple learning settings considering various labeling ratios, which unleashes the potential of PGFL.

In recent years, Graph Neural Networks (GNNs) have attracted substantial attention due to their powerful ability in modeling graph-structured data and broad applications across various domains such as social media, biology, health and finance. Despite these successes, GNNs exhibit significant vulnerabilities to adversarial attacks, which poses challenges to their reliable deployment in real scenarios. In this tutorial, we will provide an in-depth exploration of existing adversarial attacks and the state-of-the-art techniques in enhancing the robustness of GNNs. Participants will gain insights into advancing attack and defense methods, along with evaluation and comparison of robust GNN models. Besides a thorough overview on current landscape, we will also cover the summary and discussion on potential future directions, aiming to inspire more researchers to engage and innovate in this field.

To improve the handling performance of four-in-wheel-motor-drive (4IWMD) and four-wheel-steering (4WS) vehicles, an integrated control scheme based on active rear steering (ARS) and torque coordination is developed in this study. Considering the tracking of reference states, and the constraints of motor torque and steering angle, the integrated control scheme is designed through model predictive control (MPC). The active rear steering and direct yaw moment generated by the in-wheel motors can assist the vehicle in tracking the desired yaw rate to improve the handling performance. By utilizing the anti-dive forces of the vehicle suspensions, roll motion can be directly controlled via torque coordination without using active suspension. Simulation on a single lane change maneuver is performed, and the results demonstrate that the proposed controller can effectively improve the handling performance and ensure roll stability.

Implicit models such as Deep Equilibrium Models (DEQs) have emerged as promising alternative approaches for building deep neural networks. Their certified robustness has gained increasing research attention due to security concerns. Existing certified defenses for DEQs employing interval bound propagation and Lipschitz-bounds not only offer conservative certification bounds but also are restricted to specific forms of DEQs. In this paper, we provide the first randomized smoothing certified defense for DEQs to solve these limitations. Our study reveals that simply applying randomized smoothing to certify DEQs provides certified robustness generalized to large-scale datasets but incurs extremely expensive computation costs. To reduce computational redundancy, we propose a novel Serialized Randomized Smoothing (SRS) approach that leverages historical information. Additionally, we derive a new certified radius estimation for SRS to theoretically ensure the correctness of our algorithm. Extensive experiments and ablation studies on image recognition demonstrate that our algorithm can significantly accelerate the certification of DEQs by up to 7x almost without sacrificing the certified accuracy. The implementation will be publicly available upon the acceptance of this work. Our code is available at https://github.com/WeizhiGao/Serialized-Randomized-Smoothing.

The adversarial robustness of Graph Neural Networks (GNNs) has been questioned due to the false sense of security uncovered by strong adaptive attacks despite the existence of numerous defenses. In this work, we delve into the robustness analysis of representative robust GNNs and provide a unified robust estimation point of view to understand their robustness and limitations. Our novel analysis of estimation bias motivates the design of a robust and unbiased graph signal estimator. We then develop an efficient Quasi-Newton Iterative Reweighted Least Squares algorithm to solve the estimation problem, which is unfolded as robust unbiased aggregation layers in GNNs with theoretical guarantees. Our comprehensive experiments confirm the strong robustness of our proposed model under various scenarios, and the ablation study provides a deep understanding of its advantages.

Transformer-based architectures have dominated various areas of machine learning in recent years. In this paper, we introduce a novel robust attention mechanism designed to enhance the resilience of transformer-based architectures. Crucially, this technique can be integrated into existing transformers as a plug-and-play layer, improving their robustness without the need for additional training or fine-tuning. Through comprehensive experiments and ablation studies, we demonstrate that our ProTransformer significantly enhances the robustness of transformer models across a variety of prediction tasks, attack mechanisms, backbone architectures, and data domains. Notably, without further fine-tuning, the ProTransformer consistently improves the performance of vanilla transformers by 19.5\%, 28.3\%, 16.1\%, and 11.4\% for BERT, ALBERT, DistilBERT, and RoBERTa, respectively, under the classical TextFooler attack. Furthermore, ProTransformer shows promising resilience in large language models (LLMs) against prompting-based attacks, improving the performance of T5 and LLaMA by 24.8\% and 17.8\%, respectively, and enhancing Vicuna by an average of 10.4\% against the Jailbreaking attack. Beyond the language domain, ProTransformer also demonstrates outstanding robustness in both vision and graph domains.

Single-cell assay for transposase-accessible chromatin using sequencing (scATAC-seq) data provided new insights into the understanding of epigenetic heterogeneity and transcriptional regulation. With the increasing abundance of dataset resources, there is an urgent need to extract more useful information through high-quality data analysis methods specifically designed for scATAC-seq. However, analyzing scATAC-seq data poses challenges due to its near binarization, high sparsity and ultra-high dimensionality properties. Here, we proposed a novel network diffusion-based computational method to comprehensively analyze scATAC-seq data, named Single-Cell ATAC-seq Analysis via Network Refinement with Peaks Location Information (SCARP). SCARP formulates the Network Refinement diffusion method under the graph theory framework to aggregate information from different network orders, effectively compensating for missing signals in the scATAC-seq data. By incorporating distance information between adjacent peaks on the genome, SCARP also contributes to depicting the co-accessibility of peaks. These two innovations empower SCARP to obtain lower-dimensional representations for both cells and peaks more effectively. We have demonstrated through sufficient experiments that SCARP facilitated superior analyses of scATAC-seq data. Specifically, SCARP exhibited outstanding cell clustering performance, enabling better elucidation of cell heterogeneity and the discovery of new biologically significant cell subpopulations. Additionally, SCARP was also instrumental in portraying co-accessibility relationships of accessible regions and providing new insight into transcriptional regulation. Consequently, SCARP identified genes that were involved in key Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways related to diseases and predicted reliable cis-regulatory interactions. To sum up, our studies suggested that SCARP is a promising tool to comprehensively analyze the scATAC-seq data.

Learning to represent and simulate the dynamics of physical systems is a crucial yet challenging task. Existing equivariant Graph Neural Network (GNN) based methods have encapsulated the symmetry of physics, \emph{e.g.}, translations, rotations, etc, leading to better generalization ability. Nevertheless, their frame-to-frame formulation of the task overlooks the non-Markov property mainly incurred by unobserved dynamics in the environment. In this paper, we reformulate dynamics simulation as a spatio-temporal prediction task, by employing the trajectory in the past period to recover the Non-Markovian interactions. We propose Equivariant Spatio-Temporal Attentive Graph Networks (ESTAG), an equivariant version of spatio-temporal GNNs, to fulfil our purpose. At its core, we design a novel Equivariant Discrete Fourier Transform (EDFT) to extract periodic patterns from the history frames, and then construct an Equivariant Spatial Module (ESM) to accomplish spatial message passing, and an Equivariant Temporal Module (ETM) with the forward attention and equivariant pooling mechanisms to aggregate temporal message. We evaluate our model on three real datasets corresponding to the molecular-, protein- and macro-level. Experimental results verify the effectiveness of ESTAG compared to typical spatio-temporal GNNs and equivariant GNNs.

BACKGROUND: In the field of biology and medicine, the interpretability and accuracy are both important when designing predictive models. The interpretability of many machine learning models such as neural networks is still a challenge. Recently, many researchers utilized prior information such as biological pathways to develop neural networks-based methods, so as to provide some insights and interpretability for the models. However, the prior biological knowledge may be incomplete and there still exists some unknown information to be explored. RESULTS: We proposed a novel method, named PathExpSurv, to gain an insight into the black-box model of neural network for cancer survival analysis. We demonstrated that PathExpSurv could not only incorporate the known prior information into the model, but also explore the unknown possible expansion to the existing pathways. We performed downstream analyses based on the expanded pathways and successfully identified some key genes associated with the diseases and original pathways. CONCLUSIONS: Our proposed PathExpSurv is a novel, effective and interpretable method for survival analysis. It has great utility and value in medical diagnosis and offers a promising framework for biological research.