Search for: All records

Electroencephalogram (EEG) signals are vital for automated seizure detection, but their inherent noise makes robust representation learning challenging. Existing graph construction methods, whether correlation-based or learning-based, often generate redundant or irrelevant edges due to the noisy nature of EEG data. This significantly impairs the quality of graph representation and limits downstream task performance. Motivated by the remarkable reasoning and contextual understanding capabilities of large language models (LLMs), we explore the idea of using LLMs as graph edge refiners. Specifically, we propose a two-stage framework: we first verify that LLMbased edge refinement can effectively identify and remove redundant connections, leading to significant improvements in seizure detection accuracy and more meaningful graph structures. Building on this insight, we further develop a robust solution where the initial graph is constructed using a Transformer-based edge predictor and multilayer perceptron, assigning probability scores to potential edges and applying a threshold to determine their existence. The LLM then acts as an edge set refiner, making informed decisions based on both textual and statistical features of node pairs to validate the remaining connections. Extensive experiments on TUSZ dataset demonstrate that our LLM-refined graph learning framework not only enhances task performance but also yields cleaner and more interpretable graph representations. 

Seizure detection based on EEG signals is highly challenging due to complex spatiotemporal dynamics and extreme inter-patient variability. To model such complex patterns, recent methods construct dynamic graphs via statistical correlations, predefined similarity measures, or implicit learning, yet rarely account for EEG's highly noisy nature. Consequently, these graphs usually contain redundant or task-irrelevant connections, undermining model performance even when using state-of-the-art architectures. In this paper, we present a new perspective for EEG seizure detection: jointly learning denoised dynamic graph structures and informative spatial-temporal representations guided by the Information Bottleneck (IB). Unlike prior approaches, our graph constructor explicitly accounts for the noisy characteristics of EEG data, producing compact and reliable connectivity patterns that better support downstream seizure detection. To further enhance representation learning, we employ a self-supervised Graph Masked AutoEncoder that reconstructs masked EEG signals based on dynamic graph context, promoting structure-aware and compact representations that align with the IB principle. Bringing things together, we introduce Information Bottleneck-guided EEG SeizuRE DetectioN via SElf-Supervised Learning (IRENE), which explicitly learns dynamic graph structures and interpretable spatial-temporal EEG representations. IRENE addresses three core challenges: (i) Identifying the most informative nodes and edges; (ii) Explaining seizure propagation in the brain network; and (iii) Enhancing robustness against label scarcity and inter-patient variability. Extensive experiments on benchmark EEG datasets demonstrate that our method outperforms state-of-the-art baselines in seizure detection and provides clinically meaningful insights into seizure dynamics. The source code is available at https://github.com/LabRAI/IRENE. 

Navigating complex real-world environments requires understanding the semantic context and effectively making decisions. Existing solutions leave room for improvements: traditional reactive approaches that do not maintain a map often struggle in complex environments, map-dependent methods demand significant effort in mapping processes, and learning-based methods rely on large training datasets and face the difficulty of generalization. To address these challenges, we propose a novel visual semantic navigation framework that combines data-driven semantic understanding, Pareto-optimal decision-making, and image-space planning. Our approach uses a local environmental representation callednavigability image, which allows the robot to assess immediate traversability without relying onapriorimapping or navigation data. Building on this, we introducePareto-Optimal Visual Navigation(POVNav), a decision-making framework in the image space that identifies appropriate subgoals, constructs collision-free paths, and generates control commands using visual servoing. This framework also supports selective navigation behaviors, such as avoiding traversable yet slippery grasslands to prevent getting stuck, by dynamically adjusting the navigability criteria within the local representation. POVNav is lightweight, operating solely with a monocular camera and without requiring map storage or training data collection, making it highly versatile for different robotic platforms and environments. Extensive year-round real-world experiments validated its efficacy in both structured indoor environments and unstructured outdoor settings, including dense forest trails and snow-covered roads. Field experiments using various image segmentation techniques demonstrated its robustness and adaptability across a wide range of conditions. Additionally, we demonstrate that POVNav successfully guides a robot through narrow pipes in a culvert inspection task. Overall, we showcase the utility of POVNav in real-world scenarios, highlighting its flexibility and computational efficiency for autonomous robots in complex environments. 

Accurately predicting the oxidative stability of battery electrolytes is crucial for improving our understanding of high-voltage behavior and rational design of next-generation systems employing novel chemistries. However, commonly applied strategies based on evaluation of orbital occupancies of isolated molecules within density functional theory techniques neglect many-body solvation and interfacial effects that govern the electro-thermodynamics in real systems. Here, we advance a computational methodology that integrates molecular dynamics sampling of local solvation environments with explicit vertical ionization potential (IP) calculations to account for such effects. Our approach allows for both statistical accounting of IP distributions as well as prediction of the oxidized species (e.g., solvent vs anion decomposition). Application of this method to a matrix of electrolytes based on common lithium salts and solvents yields more detailed conclusions that often disagree with those gained through conventional calculations. We also demonstrate that this methodology can capture variations in IP associated with increased salt concentrations as well as the speciation and stability next to electrified model interfaces. This work offers a comprehensive accounting of the microscopic factors and electronic structure considerations that stabilize molecules and their unique solvation environment in modern electrochemical systems. 

The physical and chemical properties of electrolytes have significant impacts on battery performance. The concept of nanoconfinement has been proposed as an innovative modification strategy to address challenges related to the thermal stability, ion transport efficiency, and electrochemical stability of electrolytes. This involves confining electrolytes within nanoscale or sub-nanoscale spaces, leading to improvements in their physicochemical properties, such as increased boiling points, optimized ion migration, regulated ion concentration gradients, effective ion sieving, accelerated charge transfer, and suppressed side reactions. In this perspective article, we highlight the substantial potential of these approaches for extending the cycle life, broadening operational conditions, and enhancing the safety of lithium-based batteries. Additionally, the challenges and future research directions in this area are discussed. 

Abstract Sodium all‐solid‐state batteries (NaSSBs) with an alloy‐type anode (e.g., Sn and Sb) offer superior capacity and energy density compared to hard carbon anode. However, the irreversible loss of Na⁺at the alloy anode during the initial cycle results in diminished capacity and stability, impairing full‐cell performance. This study presents an easy‐to‐implement cathode presodiation strategy by employing a Na‐rich material to address these challenges. Leveraging the high theoretical capacity and suitable voltage window, Na₂S is chosen as the Na donor, which is activated by creating a mixed electron‐ion conducting network, delivering a high capacity of 511.7 mAh g⁻¹. By adding a small amount (i.e., 3 wt.%) of Na₂S to the cathode composite, a NaCrO₂|| Sn full cell demonstrated capacity improvement from 90.8 to 118.2 mAh g⁻¹(based on cathode mass). The capacity‐balanced full cell can thus cycle to more than 300 times with >90% capacity retention. This work provides a practical solution to enhance the full‐cell performance and advance the transformation from half‐cell to full‐cell applications of NaSSBs. 

Abstract Matching the capacity of the anode and cathode is essential for maximizing electrochemical cell performance. This study presents two strategies to balance the electrode utilization in zinc ion supercapacitors, by decreasing dendritic loss in the zinc anode while increasing the capacity of the activated carbon cathode. The anode current collector was modified with copper nanoparticles to direct zinc plating orientation and minimize dendrite formation, improving the Coulombic efficiency and cycle life. The cathode was activated by an electrolyte reaction to increase its porosity and gravimetric capacity. The full cell delivered a specific energy of 192 ± 0.56 Wh kg⁻¹at a specific power of 1.4 kW kg⁻¹, maintaining 84% capacity after 50,000 full charge-discharge cycles up to 2 V. With a cumulative capacity of 19.8 Ah cm⁻²surpassing zinc ion batteries, this device design is particularly promising for high-endurance applications, including un-interruptible power supplies and energy-harvesting systems that demand frequent cycling.