The core challenge of Long-Term Time Series Forecasting (LTSF) lies in effectively decoupling and modeling the intertwined long-term dependencies and high-frequency short-term perturbations within the data. In recent years, architectures represented by MLP-Mixer have shown great potential in capturing the global, long-range dependencies of sequences. However, their inherent global information mixing mechanism has limited ability to perceive local temporal structures and instantaneous dynamics. To address this limitation, we propose WaveMixerNet, a dual-stream architecture that utilizes wavelet decomposition to explicitly separate and model trend and detail information. Our model first employs a discrete wavelet transform to decompose the input sequence into a low-frequency approximation component and a high-frequency detail component. These components are then fed into two specialized streams for processing: one is the Approximation Feature Extractor (AFE), built upon MLP-Mixer, dedicated to processing the low-frequency component to effectively model the core long-term trend of the sequence; the other is the Detail Feature Extractor (DFE), based on a Convolutional Neural Network (CNN), responsible for processing the high-frequency component to capture instantaneous dynamics and local patterns. By systematically fusing the long-term trend features extracted by the AFE with the short-term detail features extracted by the DFE, WaveMixerNet constructs a comprehensive representation of the time series. This explicit separation and specialized modeling of trends and details greatly enhance the model's long-term forecasting capabilities. Our model achieves strong performance and ranks among the leading methods on multiple datasets. The code is available at https://github.com/wang976/WaveMixerNet.
As recent Multi-Layer Perceptron (MLP) mixer models have achieved state-of-the-art performance in time series forecasting, modeling each MLP-mixer as a separate expert within a mixture is expected to extend the representational capacity of the model, allowing each expert to be activated in response to time-varying inputs. However, extending MLP-mixers into a Mixture-of-Experts (MoE) architecture introduces a significant increase in the number of trainable parameters, rendering the model more challenging to train. To mitigate this problem, we propose a method that composes a fully trainable global expert and multiple non-trainable local experts. Specifically, our approach clones the weights of the global expert into the local experts and then modifies their weight distributions using moment learning, a recently proposed unconventional method for training neural networks. Concretely, each local expert is produced by applying moment-based transformations to a shared copy of the global expert's weights, so that expert specialization is obtained without independently training the additional experts. Experimental results using a lightweight Time Series Mixer (TSMixer) architecture demonstrate that our method achieves performance competitive with fully trainable MoE counterparts, without introducing a significant increase in trainable parameters. Across multiple benchmark settings, the proposed model attains forecasting accuracy on par with, and in several cases favorable to, a fully trainable multi-expert baseline while adding only a small fraction of the extra trainable parameters that such a baseline requires, and this efficiency is further corroborated by measurements of memory footprint as well as an effect-size-based assessment of the observed differences.
The effective creation of better, faster, and more efficient electronic circuits particularly in the radio frequency (RF) segment, is more important than ever for the entire world. A prevalent issue in wireless communication systems, such as those used for charging or Vehicle-to-Everything (V2X) communication, interference and signal reception problems are related to the receiver front end of electric vehicles (EV) applications. Modern receiver technologies are one way to address this issue, as they reduce interference and enhance signal reception. To address the above requirements, this paper introduces a down conversion Gilbert mixer with the improvement in conversion gain, power efficiency, linearity and noise performance. Here, the proposed mixer design uses current source helpers and current bleeding technique to enhance the performance of the conventional Gilbert mixer. The proposed mixer is simulated in Cadence Tool with an intermediate frequency (IF) of 100 MHz, which provides a conversion gain of 12 dB with the third order input intercept point (IIP3) of 4.5 dBm and a noise figure (NF) of about 9.86 dB. Also, the design functions at a supply voltage of 1.2 V, with a power consumption of only 0.96 mW. So, this mixer design may be the correct choice as a core component in the receiver front end in applications like HEVs and sensor networks, especially where wireless communication, signal conversion, and sensing are involved.
The increasing adoption of continuous-flow chemistry has shifted attention toward transport-controlled phenomena, particularly mixing, as critical determinants of reaction performance. Mixing time is a fundamental yet often ill-defined parameter in flow chemistry, particularly under turbulent, high-throughput conditions where mixing is commonly assumed rather than directly measured. Here, we demonstrate direct measurement of the mixing time in a turbulent T-junction micromixer by using chemiluminescence as a real-time mixing clock that reports hydrodynamic homogenization. Because the intrinsic reaction time of the luminol chemiluminescence system is orders of magnitude shorter than the residence time, the temporal evolution of light emission directly reflects the progression of scalar homogenization, separated from chemical kinetics or molecular diffusion. Applying this concept to a transparent T-mixer with an effective channel diameter of 500 μm, we access Reynolds numbers exceeding 6000 while maintaining pressure drops below 0.5 MPa, a regime relevant to practical flow synthesis. Under these conditions, submillisecond micromixing is directly observed, with mixing times below 0.5 ms. This micromixing time is shorter than the previously reported micromixing time of 1.3 ms in the 250 μm T-mixer. The Péclet number of the system exceeds 107, indicating that turbulent stretching-dominated homogenization drives such quick micromixing even in the large-diameter T-mixer used in flow chemistry. By experimentally defining mixing time through direct observation of homogenization, without invoking reaction models or assumed micromixing parameters, this work establishes a general framework for designing mixers and reaction lengths for better utilization of short-lived intermediates and high-throughput flow chemistry.
Phase measurement and compensation are important for high-performance optical communication systems. Previous studies have shown that most phase compensation methods rely on modifying the optical path or structure of the 90° optical hybrid mixer, which increases system complexity and cost. For highly integrated on-chip 90° optical hybrid mixers, however, such structural modification is impractical, creating a need for a flexible post-processing compensation approach. To address this, we propose a method that retrieves phase information from optical intensity measurements to determine the four output phases of the mixer. A compensation method is introduced that reconstructs the phase relationships at the mixer's outputs without modifying the chip. First, simulations verify the effectiveness of the algorithm, demonstrating that the maximum phase error is reduced from 14.62° to 6 × 10-4°. Based on this, a compensation matrix is constructed to reconstruct the signal phase relationships, thereby enhancing measurement accuracy. Experimental results confirm that the compensated phase error can be controlled within 4.21°. Compared with traditional schemes, the proposed method accomplishes phase compensation without complex optical setups, offering a practical solution for integrated photonic systems.
An optimal CMOS up-conversion mixer is designed using a novel combination of genetic algorithms and particle swarm optimization (ComGAPSO) and improved-Kolmogorov-Arnold networks (ImGKAN) for 5G communication. The proposed ImGKAN, trained with ComGAPSO, enhances optimization through social interactions and private cognition through social interactions. The proposed hybrid approach enables accurate parameter determination due to the effective modeling and compensation of nonlinearities in the up-conversion mixer. The proposed optimized mixer incorporates an enhanced linearity boosting technique (LBT) along with a tunable capacitive feedback common-source (TCF-CS) structure. This combination effectively suppresses third-order nonlinear distortion while compensating for parasitic capacitances to improve gain performance and enhance circuit stability. The proposed design achieves a peak conversion gain (CG) of approximately 4.2 dB near 24 GHz. In terms of isolation characteristics, the LO-IF isolation reaches about -44 dB. Additionally, the RF-IF isolation is around -30 dB, ensuring minimal undesired coupling between the input and output paths, while the LO-RF isolation is maintained near -39 dB. The optimized mixer exhibits an output 1 dB compression point (OP1dB) of 5.1 dBm and an input 1 dB compression point (IP1dB) of -1.1 dBm. The RF port shows a return loss of approximately -24 dB near 24 GHz. The LO port exhibits a return loss in the range of -3 to -5 dB, with improved matching observed over the operating band. Meanwhile, the IF port demonstrates strong matching at lower frequencies, with return loss values dropping below -20 dB. Furthermore, the measured optimized design achieves a minimum noise figure (NF) of approximately 3.8 dB at 24 GHz.
Long-term time series forecasting (LTSF) underpins critical applications from energy management to weather prediction, yet achieving reliable multi-step-ahead accuracy remains challenging. Existing LTSF approaches, dominated by MLP- and Transformer-based architectures, either rely on simple linear mappings or introduce increasingly complex hand-crafted inductive biases, raising the question of whether a more expressive nonlinear modeling core could offer a useful alternative. In this work, we investigate whether Kolmogorov-Arnold Networks (KANs), which use learnable basis functions on network edges to model nonlinear relationships, can serve as effective modeling components for LTSF, and under which design choices they are most useful. Motivated by this question, we propose KANMixer, a compact KAN-centered architecture consisting of a multi-scale pooling frontend, KAN-based temporal mixing blocks, and KAN-based prediction heads. Unlike KAN-based forecasting models that combine KAN with decomposition-heavy or mixture-based pipelines, KANMixer is designed as a simple and controlled architecture for examining the role of KAN components in LTSF. Under a unified five-run reproduction protocol on seven standard benchmarks, KANMixer achieves competitive performance against representative LTSF baselines, especially on ETT-style datasets, while showing dataset-dependent limitations. Additional statistical tests, ablations, efficiency profiling, Gaussian-noise evaluation, and hyperparameter sensitivity analysis show that the practical value of KAN depends on basis-function choice, architectural placement, and computational constraints. These results suggest that KANs are promising but not plug-and-play components for LTSF, and that their benefits should be evaluated together with robustness and efficiency trade-offs.
A dual-channel single-sideband mixer with high spurious suppression, high channel isolation, and wide operational bandwidth is proposed. The scheme employs a dual-polarization dual-drive Mach-Zehnder modulator (DPol-DDMZM) to simultaneously modulate two independent radio frequency (RF) signals. Carrier suppression for both RF signals with different power ratios is achieved by adjusting the polarization angle of the local oscillator (LO). Switching between up- and down-conversion within each channel is controlled by the phase difference of the corresponding RF input, allowing independent channel operation. Output channel selection is realized by tuning the parameters of a polarizer. Simulation results show that the system achieves a spurious suppression ratio exceeding 29.63 dB and channel isolation above 30 dB, demonstrating high spurious rejection and isolation. When driven by a 250 MHz 16-quadrature amplitude modulation (16-QAM) signal, the mixer achieves an adjacent channel power ratio (ACPR) up to 45.6 dB and a minimum error vector magnitude (EVM) of 0.9%. Furthermore, by adjusting the polarization controller within a ±0.4∘ range, carrier suppression better than 56 dB is maintained for two RF signals with a power ratio of up to 16 dB, demonstrating robust performance under imbalanced received signal conditions.
Nature builds functional materials through simple yet powerful processes that generate structured architecture across scales-from the lamellar patterns in seashells to the zonal organization of living tissues. Emulating such complexity in engineered systems remains challenging and often requires microfabricated components, external fields, or specialized hardware. Previously, we introduced chaotic printing as a deterministic and flow- and geometry-driven strategy for fabricating structured filaments, using static mixers embedded within extrusion printheads-primarily in the context of biofabrication. We broaden the architectural and functional scope of chaotic printing by exploring diverse static mixer designs and demonstrating its compatibility with three distinct deposition modes: wet-printing, dripping, and direct ink writing. These modalities enable the generation of material constructs with chemically and biologically relevant internal organization. We showcase examples ranging from zonally arranged mammalian cells that prefigure microtissue compartments to spatially patterned bacterial consortia composed of strict and facultative anaerobes and localized mineral precipitation within hydrogel filaments. These proof-of-concept-demonstrations underscore the potential of chaotic printing for fabricating structured soft matter where internal microarchitecture enables biologically and chemically relevant processes. This study positions chaotic printing as a modular, scalable, accessible platform for generating architected materials across fields ranging from cell culture and microbiology to functional soft materials.
To develop a rapid, scalable, and eco-friendly hydrogel for wastewater decontamination, an industrial internal mixer was employed to fabricate a green hydrogel adsorbent based on starch and humic acid. Starch-grafted polyacrylamide hydrogels incorporated with humic acid were rapidly synthesized within 5 min using an internal mixer as the reactor. Starch was gelatinized in situ, followed by graft polymerization with acrylamide via free-radical polymerization and cross-linking with N,N'-methylenebisacrylamide in the same reactor. Humic acid was introduced as a natural modifier to boost the dye adsorption performance of the starch-based hydrogel. The adsorption capacity for methylene blue was evaluated under different humic acid dosages, and the maximum adsorption capacity occurred at a humic acid dosage of 20 g. Additional batch experiments revealed that the adsorption capacity increased with increasing the solution pH (from 22.6 to 54.0 mg g-1 over pH 4-12) and the initial dye concentration (from 14.6 to 34.1 mg g-1 over 40-200 mg L-1), while it decreased with increasing the adsorbent dosage (from 35.9 to 10.6 mg g-1 over 0.1-0.9 g). Distinguished from conventional laboratory stirred reactors that require 2-3 h for hydrogel synthesis, the internal mixer achieves a one-pot synthesis within 5 min, showing outstanding potential for industrial large-scale production. This work provides a time-efficient, industrially compatible strategy to prepare eco-friendly starch-humic acid hydrogels, which show promising potential as sustainable adsorbents for dye-contaminated wastewater treatment.
With the rapid development of the animal leather market, achieving accurate identification of leather from different animal species has become a critical issue for ensuring market regulation and protecting consumer rights. Raman spectroscopy, renowned for its non-destructive, rapid, and highly sensitive nature, has been extensively applied in the qualitative and quantitative analysis of complex materials. However, practical applications often encounter challenges such as high-dimensional redundancy, noise interference, spectral peak shifts, and category overlap, hindering traditional models from achieving stable and reliable classification performance. To address these issues, this study proposes a calibration-driven multi-model consensus learning framework-MSAMixerNet. This framework integrates a multi-scale MLP-Mixer backbone network, a Transformer-based global modeling module, a CBAM channel-spatial attention mechanism, and a positional encoding mechanism, thereby balancing local sensitivity with global structural modeling capabilities. Concurrently, it introduces a Bagging multi-model ensemble strategy. By constructing multiple submodels on different training subsets and performing consensus voting, it effectively mitigates overfitting and data distribution imbalance issues. Experimental results on animal leather Raman spectroscopy classification demonstrate that this framework achieves an accuracy of 98.27% across nine animal leather datasets, outperforming existing mainstream methods in both robustness and confidence reliability. This research not only provides a novel solution for intelligent spectral recognition in high-dimensional noisy environments but also offers a feasible technical pathway for subsequent market regulation and material anti-counterfeiting applications.
We have developed a methodology for rapid extraction under continuous-flow conditions. Basic compounds with a tert-butoxycarbonyl protecting group, which are labile under acidic conditions, were extracted into an acidic aqueous phase by a mixer-settler under continuous-flow conditions and immediately neutralized prior to their decomposition. This methodology could serve as a promising option for purifying drug substances during pharmaceutical production.
The reoriented potential mixer (RPM) can enhance mixing and reaction during in situ remediation of contaminated groundwater, in which a chemical or biological amendment is introduced into a contaminant plume to react with and degrade the contaminant. Each step of the RPM consists of dipole flow between a pair of wells-one for injection and one for extraction-that is active for a dimensionless time τ. Subsequently, a different pair of wells is activated at a reorientation angle Θ. Certain {τ,Θ} pairs lead to chaotic advection, with elliptic islands embedded in a chaotic sea. Importantly, a central elliptic island encompasses a "trapped" region containing fluid that does not reach an extraction well during typical groundwater remediation timescales. Indeed for groundwater remediation, the RPM must be designed so that the contaminant remains within such a trapped region, eliminating the possibility of degrading previously uncontaminated regions of the aquifer. In our model, this trapped region consists of two parts: an inner region surrounding an elliptic fixed point within which the fluid is minimally mixed by weak shear, and an outer region with stronger shear that can produce substantial stretching of the interface between the contaminant and amendment plumes. An RPM designed to contain the contaminant and amendment plumes within the trapped region has the potential to enhance reaction during in situ remediation of contaminated groundwater, while preventing the spread of contaminated groundwater into previously uncontaminated regions of the aquifer.
Stereo matching has become an increasingly important component of modern autonomous systems. Developing deep learning-based stereo-matching models that deliver high accuracy while operating in real time continues to be a major challenge in computer vision. In the domain of cost volume-based stereo matching, accurate disparity estimation depends heavily on large-scale cost volumes. However, such large volumes store substantial redundant information and also require computationally intensive aggregation units for processing and regression, making real-time performance unattainable. Conversely, small-scale cost volumes followed by lightweight aggregation units provide a promising route for real-time performance, but lack sufficient information to ensure highly accurate disparity estimation. To address this challenge, we propose the Enhanced Shuffle Mixer (ESM) to mitigate information loss associated with small-scale cost volumes. ESM restores critical details by integrating primary features into the disparity upsampling unit. It quickly extracts features from the initial disparity estimation and fuses them with image features. These features are mixed by shuffling and layer splitting, then refined through a compact feature-guided hourglass network to recover more detailed scene geometry. The ESM focuses on local contextual connectivity with a large receptive field and low computational cost, leading to improved disparity estimation accuracy while maintaining real-time performance under the evaluated settings. The compact version of ESMStereo achieves an inference speed of 116 FPS on RTX 4070S and 91 FPS on the AGX Orin.
Efficient mixing in microfluidic systems is bottlenecked by laminar flow and diffusion-dominated transport, limiting the performance of biochemical assays such as conversion rates of enzymatic processes. Here, we present an acoustofluidic micromixer based on two-photon polymerization (2PP)-printed three-dimensional microneedles that act as ultrasonic microresonators to generate localized acoustic streaming. The freeform arrangement of the microneedles enables integration into diverse microfluidic channel layouts and produces strong, tunable streaming vortices that enhance transport. We characterize the streaming flow fields as a function of actuation frequency, and demonstrate rapid and voltage-dependent mixing with a 2.5-fold enhancement. When applied to the alkaline phosphatase-fluorescein diphosphate (ALP-FDP) reaction, acoustic mixing doubles the fluorescein signal under optimal conditions and significantly reduces the assay time and channel footprint required in continuous flow. These resultsdemonstrate that 3D microneedle-based acoustofluidics is a compact, versatile, and integrable solution for enhancing transport-limited biochemical processes in lab-on-a-chip systems.
暂无摘要（点击查看详情）
The development of Co-free, Ni-rich cathodes is critical to overcoming cost and sustainability challenges in lithium-ion batteries. Here, we propose a high-entropy doping strategy for LiNiO2 by uniformly incorporating Mn, Al, Fe, Mg, and Cr into the transition-metal layer through a scalable 3D mixing process. Each dopant plays a complementary role in stabilizing the layered structure and enhancing electrochemical performance. The resulting Li1.05Ni(MnAlFeMgCr)0.10O2 suppresses Li/Ni cation mixing, strengthens TM─O bonding, mitigates polarization and diffusion bottlenecks, and reduces electrode swelling (from 23% in pristine LNO to 16%). The high-entropy-doped cathode delivered a capacity retention of 79% after 100 charge-discharge cycles at 0.3 C, markedly surpassing that of pristine LNO (51%). These results demonstrate that high-entropy doping effectively improves structural robustness and long-term durability, offering a practical pathway to economically viable and sustainable next-generation Co-free Ni-rich layered cathodes.
The P300 speller is a widely adopted brain computer interface (BCI) paradigm that enables hands free character selection based on event-related potentials elicited through an oddball stimulus paradigm. Despite its utility, the system's performance is often constrained by the low signal-to-noise ratio and complex spatiotemporal characteristics of EEG signals, especially when only a limited number of repetitions or labeled samples are available. Moreover, substantial within-session calibration is typically required to achieve reliable decoding before online spelling, posing a major practical barrier. To tackle these challenges, we propose SCL-EEGMixer, a lightweight, end to-end neural architecture that combines a convolutional mixer network with supervised contrastive learning. The model extracts discriminative spatiotemporal representations via the convolutional mixer and enhances learning with a hybrid loss that fuses cross-entropy and supervised contrastive objectives. This design promotes intra-class compactness and inter-class separability, enabling robust learning from scarce labeled data. Extensive evaluations on both a public benchmark and a self-collected dataset demonstrate that SCL-EEGMixer consistently outperforms representative baselines in both binary P300 classification and character recognition tasks under a within-session protocol. Notably, it maintains high accuracy and information transfer rate even when trained with as few as one or two calibration characters, highlighting its potential for reducing within-session calibration burden in P300 spelling.
Rare earth elements (REEs) are predominantly separated by using mixer-settler solvent extraction technology. While this method has been utilized in the industry, it does have a few setbacks, such as requirement of large module footprint to compensate for inefficient separation. Likewise, researchers have turned to membrane separation technology to address these challenges faced by mixer-settler solvent extraction. Supported liquid membranes (SLMs) are a viable alternative in membrane separation technology as they exhibit similar conceptual designs to that of solvent extraction, albeit requiring smaller amounts of organic carrier and lower module footprint compared to mixer settler solvent extraction. In this research, the application of an electrospun membrane was explored using SLMs for the separation between light REEs (LREEs) and heavy REEs (HREEs). To do this, the study initially employs the usage of neodymium (Nd) and dysprosium (Dy) as the baseline for LREEs and HREEs, and later using real REE leachate for the separation between LREEs and HREEs. This study also investigates the effects of different pH of feed solution, organic carrier loading, and stripping concentrations. Overall, to achieve higher selectivity of Nd/Dy, a higher pH of feed solution of 5, a low loading of organic carrier at 10 wt%, and a low stripping concentration of 1 M H2SO4 was recommended. For the separation of LREEs and HREEs, a separation factor of 2.38 was achieved with a recovery of 10.43% of LREEs in under 60 min by using a small-scale membrane of 4 cm2 effective area. Long term operation indicates a stable LREE/HREE separation performance using the optimized condition.
Polycystic ovary syndrome (PCOS) frequently co-occurs with obesity, insulin resistance, dyslipidaemia and metabolic syndrome (MetS). However, whether this metabolic burden represents secondary comorbidity or an intrinsic component of PCOS genetic architecture remains unclear. We used genomic structural equation modelling to construct female- and male-specific hierarchical MetS latent factors from six sex-stratified cardiometabolic GWAS, including BMI, waist circumference, triglycerides, HDL cholesterol, type 2 diabetes and fasting glucose. Sex-specific MetS genetic architecture was evaluated using latent factor GWAS, sex-difference testing, MiXeR, MAGMA and GSA-MiXeR. To assess the relationship between MetS liability and PCOS, we performed CPASSOC, conjunctional FDR, Bayesian colocalization, GWAS-by-subtraction and GenomicSEM-based mediation analysis. Female and male MetS factors shared a highly overlapping polygenic backbone, but their genetic correlation was significantly below unity. Among comparable lead variants, 9.1% showed sex-dimorphic effects, and MiXeR estimated an approximately fourfold larger female-specific causal component than the male-specific component. Cross-trait analysis identified 45 candidate pleiotropic loci linking PCOS with the female MetS factor, 34 conjFDR-supported shared loci and 10 loci with strong Bayesian colocalization evidence, including FTO, ERBB3, FGFR1, KLF16 and TEX41. GWAS-by-subtraction showed that 16.7% of PCOS genetic variance was shared with female MetS, whereas 83.3% remained MetS-independent and retained stronger reproductive-endocrine features. Variant-level mediation further showed that MetS-mediated effects were locus-specific rather than a global amplifier of PCOS risk. This study redefines the metabolic comorbidity of PCOS as a separable, female-biased genetic dimension embedded within, but distinct from, the predominant reproductive-endocrine genetic architecture of PCOS. These findings provide a genetic framework for resolving reproductive-metabolic heterogeneity in PCOS and may inform future risk stratification and precision management of metabolic complications in women with PCOS.