The rapid development of multimodal epidermal sensing requires scalable, energy-efficient data processing architectures capable of processing large volumes of raw data. Conventional systems suffer from high energy consumption and transmission latency due to the physical separation of sensors and processors. Here, we present an ultrathin flexible edge computing circuit based on carbon nanotube thin-film transistors (CNT-TFTs) and machine learning (ML)-assisted design. By incorporating substrate engineering, ML-derived device modeling, and industry-compatible design methodologies, we establish a complete toolchain from device to system. The ML model achieves 91.2% prediction accuracy, enabling simulation-guided optimization of logic gates. A CNT-based standard cell library enables the construction of flexible circuits with 361 transistors and 160 logic gates. Monolithic integration with an 8-channel tilt sensor achieves 62.5% data compression while maintaining functionality after undergoing 360° deformation. This work establishes an ML-assisted CNT circuit design framework for fully integrated flexible edge computing, enabling scalable wearable applications.
Test-time adaptation (TTA) aims to adapt the model trained on source domain to unseen target domain using a few unlabeled images during inference, which holds great value for the deployment of models in the clinical practice. In this setting, the model can only access online unlabeled test samples and pre-trained model on the source domain. Because unlabeled test samples may arrive sequentially, the model needs to adjust online for the cross-domain distribution shift from different medical institutions, the scale of which would change concurrently and continually over time. However, unstable optimization and abnormal distribution will lead to error accumulation and catastrophic forgetting. Considering the role of brain extracellular space in balancing neural homeostasis and signal transmission, we recognize that the existing TTA methods lack a dedicated component to ensure the stability and accuracy of the model. In this paper, we propose a robust TTA approach for cross-domain segmentation as MemTTA. Specifically, firstly, we introduce transductive batch normalization to ensure stability, which calculates the mean and the variance from the source domain and current test batch. Secondly, we propose a memorized spatial pixel-level clustering strategy to represent each category with multiple and anisotropic prototypes for feature alignment, which can be associated with the parametric classifier. During test time, we adapt the segmentation model to each test batch with self-supervision augmentation consistency learning to improve the inference performance. MemTTA needs only one epoch training on each test batch, and then is comparable to standard models as the traditional inference pipeline. The proposed method is extensively evaluated on neuron, brain metastases, cardiac, and abdominal organ image segmentation. The experimental results demonstrate that our proposed MemTTA can effectively mitigate test-time domain shift and catastrophic forgetting, and is superior to existing state-of-the-art approaches.
Shortly after the failed PKI uprisings of 1926/27, Tan Malaka and his associates established the Partai Republik Indonesia (PARI). Although he acted as the party chairman and chief strategist, his involvement in the party operation was minimal as he lived in exile. Nevertheless, he loomed large in the eyes of both his followers and enemies. Not only was Tan Malaka a legendary guru for Indonesian revolutionaries, but also an enormous threat to colonial authorities across East and Southeast Asia. This chapter explores Tan Malaka's exile in China between 1927 and 1936 and how such experiences reflect his shifting relationship with Indonesia's ongoing struggles for independence, the international communist movement, and the surveillance and policing practices of multiple colonial states.
Microbial functions and metabolism are intrinsic drivers of pollutant removal in mixotrophic denitrification systems. Four pyrite-based mixotrophic denitrifying biofilters were constructed and monitored for 304 days. Variations in pollutant characteristics indicated that the hot zones of heterotrophic denitrification, autotrophic denitrification, and sulfate reduction were located in the bottom, middle-lower, and upper parts of biofilters, respectively. These hot zones corresponded to preferential enrichment of heterotrophic denitrifying, S-based mixotrophic denitrifying, and sulfate-reducing bacteria, respectively, highlighting microbial spatial stratification. Differential functional gene analysis for S reduction revealed that only a dissimilated sulfate reduction process could consistently provide biogenic S0 as a new electron donor via the flavocytochrome c sulfide dehydrogenase (Fcc) enzyme and extracellular polymeric substance protection systems, enhancing the denitrification process. X-ray photoelectron spectroscopy confirmed the accumulation of biogenic S0 . Untargeted metabolomic analysis suggested that vitamin B12 and tryptophan might be the key metabolites for realizing synergistic promotion of autotrophic and heterotrophic denitrification. The microbe-metabolite network indicated that dominant bacteria (e.g., Thiothrix and unclassified\_f\_Rhodocyclaceae) were specialists with less cross-feeding metabolism, while rare species (e.g., Thiobacillus and Desulfobacter) were generalists with complex cross-feeding metabolism in the constructed mixotrophic denitrification systems. The electron transfer pattern indicated that most of the electrons released from S, C, and Fe oxidation were utilized in denitrification processes as the dominant nitrogen removal pathway, including S2- /S0-based autotrophic, fermentation acetic acid production-heterotrophic, and Fe(II)-based autotrophic denitrification. Some electrons were utilized for coupling dissimilatory nitrate reduction to ammonia (DNRA) and anammox processes as an auxiliary pathway for systemic nitrogen removal. The findings of this study advance our understanding of the deeper intrinsic drivers of nitrogen removal by pyrite-based mixotrophic denitrifying biofilters, facilitating their optimization. (c) 2025 THE AUTHORS. Published by Elsevier LTD on behalf of Chinese Academy of Engineering and Higher Education Press Limited Company. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Nanoconfined water—ubiquitous across both engineered nanoporous adsorbents and subsurface geological formations—plays a pivotal yet underexplored role in carbon capture and storage (CCS). This review systematically examines the physicochemical properties and functional implications of water confined within nanoporous environments, emphasizing its dualistic impact on both CO2 capture and geological CO2 storage. We first summarize recent advances from computational simulations and experimental characterizations, highlighting the altered thermodynamic and structural features, dynamic behavior, dielectric properties, and chemical reactivity of nanoconfined water. We then integrate insights from surface chemistry, materials science, and geoscience to elucidate how nanoconfined water influences CCS processes through competitive adsorption, pore accessibility, wettability, solubility, and mineralization kinetics, spanning systems from nanoporous adsorbents such as zeolites, metal–organic frameworks (MOFs), and activated carbon (AC) to unconventional formations including shale and tight sandstone. These findings also suggest opportunities for practical applications, such as guiding the design of hydrophobic MOFs for improved CO2 capture and supporting strategies to preserve caprock integrity in subsurface storage. Finally, we identify key challenges in bridging molecular-level understanding with material- and reservoir-scale performance, emphasizing the need for multiscale experimental techniques, realistic molecular modeling, and cross-disciplinary strategies to fully harness the functional potential of nanoconfined water in CCS.
Ice crystal orientation strongly influences cloud radiative properties and remote sensing retrievals, but long-term, high-resolution quantitative observations remain scarce. This study presents comprehensive case studies and statistical analyses of horizontally oriented ice crystals (HOICs) based on full-year (2022) synergistic observations in Beijing, China, combining a zenith-pointing micropulse lidar, a collocated 15° off-zenith polarization lidar, and a Ka-band cloud Doppler radar. Applying a novel height-resolved classification method based on dual-angle polarization lidars, HOICs are identified with high spatiotemporal resolution. HOICs are found to be common, accounting for 15.0 % of all ice-containing cloud data points annually with a peak of 24.6 % in summer, and showing maximum occurrence at temperatures from −20 to −10 °C. Macroscopically, typical HOIC layers exhibit horizontal extents of 10–100 km and durations of several hours within their optimal formation temperature ranges. Furthermore, the Euclidean-distance analysis between HOICs and past overlying cloud layers reveals a strong linkage of HOIC occurrence to supercooled liquid water clouds (SWCs), being much closer to HOIC events than randomly oriented ice crystals (ROICs). Dynamically, cloud radar observations further reveal that HOICs preferentially occur in stable environments with turbulent eddy dissipation rates below 10−2 m2 s−3 and exhibit lower fall velocities than ROICs. Estimates based on radar-observed vertical velocity indicate typical HOIC equivalent diameters of approximately 1200 µm with Reynolds numbers predominantly below 100. These findings provide key observational constraints for improving ice cloud microphysics and orientation parameterizations in numerical models.