Low-pass Research Articles

An extendable mCWSCI converter for high-gain DC-DC conversion/regulation

ABSTRACT The analysis, design, and implementation of an extendable closed-loop booster scheme, named multistage-Cockcroft-Walton-doubler switched-coupled-inductor (mCWSCI) converter, is presented and addressed for further meeting the demands on high-gain DC-to-DC power conversion/regulation. The power part of mCWSCI consists of two subgroups: switched-coupled-inductor (SCI) booster and m-stage Cockcroft-Walton doubler (CWD), working to obtain the total voltage gain [ m n + 1 + nD + 1 ] / 1 − D theoretically ( D : duty cycle, n : turns ratio). When n = 3 , m = 2 , D = 0.5 , the output can be raised to 21 times the voltage of supply (e.g. 24 V to 500 V, applied in sputters). The control part of mCWSCI consists of three subgroups: isolated low-pass filter, sawtooth wave generator, and pulse-width-modulation (PWM) block, operating together to realise the topological timing control and reinforce the output regulation/robustness. Further, detailed analysis/design is provided, including steady-state response, voltage conversion ratio (VCR), power efficiency, inductance/capacitance selection, converter stability, and control compensation. Finally, several cases are considered via testing on the software of SPICE/on the prototype of mCWSCI, and the results are illustrated not only to examine the effectiveness of the analysis/design but also to show the practicality of this presented scheme (e.g. comparison of ratios: VCR-to-component-count).

Accurate pulse-to-pulse stability measurement for femtosecond lasers

Short-term energy stability metrology of pulsed lasers is the source of widespread misleading comparisons. The most used value for this characterization in commercial products and laboratory applications is the rms pulse-to-pulse stability (P-to-P). For frequency-specific analysis, the relative intensity noise (RIN) power spectral density is measured and the integrated RIN over a certain bandwidth is given. In principle, integrated RIN and P-to-P correspond to the same value. In practice, the different experimental setups often cause results to differ significantly. In this paper, we detail the specifics of measurement setups for short-term stability characterization and focus on measurements with stability better than 1% rms. The sensitivity of the P-to-P approach with respect to the oscilloscope specifications is shown by measuring a commercial femtosecond amplifier using three different oscilloscopes. A simple modified setup, including a low-pass filter at the laser repetition rate to turn the photodetected pulse train into a quasi-sinusoidal response, is proposed. This approach allows to overcome the oscilloscope limitation in standard P-to-P setups, allowing to obtain reliable measurements lower than 1% rms with a resolution of the order of 0.1%.

A Study on Students' Perception Towards Professional Courses (CA/CS/CMA) with Reference to Coimbatore City

Abstract Professional courses such as Chartered Accountancy (CA), Company Secretaryship (CS), and Cost and Management Accountancy (CMA) are vital career pathways that offer specialized knowledge and prestigious employment opportunities. Despite their significance, these courses face declining enrolment trends and increased dropout rates, especially among students in semi-urban and urban regions. This study aims to explore the perceptions, motivations, and barriers influencing students' decision-making regarding the pursuit of CA, CS, and CMA courses in Coimbatore City. The research adopts a descriptive design and utilizes a structured questionnaire to gather primary data from higher secondary school and undergraduate students across various educational institutions in Coimbatore. The study examines factors such as awareness levels, perceived difficulty, societal and parental influence, financial constraints, availability of coaching facilities, and career aspirations. Statistical tools such as percentage analysis, chi-square test, and factor analysis are employed to interpret the data and derive meaningful insights. Findings reveal that while students recognize the value and prestige associated with professional courses, a significant proportion are deterred by factors like lengthy course duration, low pass percentages, financial burden, and lack of proper guidance. Additionally, misconceptions about the difficulty level and limited awareness about career scope also act as key deterrents. The study concludes that strategic interventions are needed to enhance awareness, simplify the learning ecosystem, and provide financial and academic support to aspiring students. Career counselling programs, early exposure to professional opportunities, and government-accredited support systems can play a pivotal role in reducing the psychological and logistical barriers faced by students. This research contributes to the understanding of educational preferences in a rapidly developing urban centre and provides insights that can aid policymakers, educational institutions, and professional bodies in making these professional courses more accessible and appealing. Key words: Professional Courses, Student Perception, Career Barriers, CA/CS/CMA, Awareness and Guidance.

Neural adaptations to temporal cues degradation in early blind: insights from envelope and fine structure vocoding

In our previous study, early-blind individuals have better speech recognition than sighted individuals, even when the spectral cue was degraded using noise-vocoders. Therefore, this study investigated the impact of temporal envelope degradation and temporal fine structure (TFS) degradation on vocoded speech recognition and cortical auditory response in early blind individuals compared to sighted individuals. The study included 20 early-blind subjects (31.20 ± 42.5 years, M: F = 11:9), and 20 age- and -sex-matched sighted subjects. Monosyllabic words were processed using the Hilbert transform to separate the envelope and TFS, generating vocoders that included only one of these components. The amplitude modulation (AM) vocoder, which contained only the envelope component, had the low-pass filter's cutoff frequency for AM extraction set at 16, 50, and 500 Hz to control the amount of AM cue. The frequency modulation (FM) vocoders, which contained only the TFS component, were adjusted to include FM cues at 50%, 75%, and 100% by modulating the noise level. A two-way repeated measures ANOVA revealed that early-blind subjects outperforming sighted subjects across almost all AM or FM-vocoded conditions (p < 0.01). Speech recognition in early-blind subjects declined more with increasing TFS degradation, as evidenced by a significant interaction between group and the degree of TFS degradation (p = 0.016). We also analyzed neural responses based on the semantic oddball paradigm using the N2 and P3b components, which occur 200–300 ms and 250–800 ms after stimulus onset, respectively. Significant correlations were observed between N2 and P3b amplitude/latency and behavioral accuracy (p < 0.05). This suggests that early-blind subjects may develop enhanced neural processing strategies for temporal cues. In particular, preserving TFS cues is considered important for the auditory rehabilitation of individuals with visual or auditory impairments.

An ultrasonic digital lock-in probing technique for monitoring fast changes within an elastic body.

A probing technique using ultrasonic waves and a digital lock-in detection scheme is developed to monitor changes within an elastic body. The scheme integrates products of the signal and reference rather than low pass filtering them. Two applications are illustrated: detecting acoustic emissions and monitoring slow dynamic nonlinear elasticity. The probe is capable of resolving the earliest times (<10-3 s) of the slow dynamic recovery and does not suffer distortion from the pump wave.

A new design of an electromyography sensor for movement detection in human-machine interaction systems

This study introduces a cost-effective, multi-channel electromyography (EMG) sensor that is capable of simultaneously acquiring electrical activity from up to four muscles. Given the susceptibility of EMG signals to noise sources like power line interference (PLI), muscle cross-talk, and skin movement artifacts, the proposed sensor is designed to enhance EMG signal fidelity. It employs fine-tuned analog filters and a Driven Right Leg (DRL) circuit, which attenuates PLI by applying an inverted and amplified common-mode voltage back into the subject’s body. The sensor uses electronic components, including AD620 instrumentation and TL074 operational amplifiers and an SMD PJ-131 audio jack. It features four independent acquisition channels, processing raw input with low-pass and high-pass filters to form a band-pass filter for signals in the 20–440 Hz, where muscle activity typically occurs. The two-layer PCB, powered by two 9-V batteries, achieves an average SNR of 23 dB. Experimental results confirm its effectiveness in upper limb movements, with significant noise suppression and clear muscle state differentiation. The DRL reference circuit improves SNR to 23 dB, compared to 20 dB with the traditional ground reference. The proposed EMG system costs around 10 euros in components, with lower PCB manufacturing costs at scale, offering a cost-effective alternative.

Identification of characteristics for RC Low Pass, High Pass, Band Pass and Band Stop Filters

In signal processing, filters are essential components used to modify the frequency content of signals, enabling the separation of desired components from unwanted noise or interference. Among the most commonly used filters in electronic systems are RC (Resistor-Capacitor) filters, which include low-pass, high-pass, band-pass, and band-stop filters. These filters are simple to implement and widely utilized for tasks such as noise reduction, signal shaping, and frequency selection. The performance of these filters varies significantly with desired frequency, and understanding these variations is crucial for their optimal application. The designed parameters and values are calculated and the gain magnitude and phase response and bode response is simulated with MATLAB.

Detection of Q235 Mild Steel Resistance Spot Welding Defects Based on EMD-SVM

Real-time detection of welding defects in resistance spot welding is a complex challenge. Dynamic resistance (DR) reflects nugget growth and varies with defect types, serving as a key indicator. This study presents an online quality evaluation and defect classification method for Q235 low-carbon steel welding. Welding current and voltage were collected in real-time, and DR signals were processed employing a second-order Butterworth low-pass filter featuring zero-phase processing to enhance accuracy. Empirical mode decomposition (EMD) decomposed these signals into intrinsic mode functions (IMFs) and residuals, which were classified by a support vector machine (SVM). Experiments showed the EMD-SVM method outperforms traditional approaches, including backpropagation (BP) neural networks, SVM, wavelet packet decomposition (WPD)-BP, WPD-SVM, and EMD-BP, in identifying four welding states: normal, spatter, false, and edge welding. This method provides an efficient, robust solution for online defect detection in resistance spot welding.

Design of an improved adaptive sliding mode observer for charge/discharge control in flywheel energy storage systems

Aiming to address severe sliding mode chattering in traditional sliding mode observer (SMO), high-frequency harmonics in the back electromotive force (EMF), and low rotor position estimation accuracy, this paper proposes an improved adaptive SMO. And considering the characteristics of the flywheel energy storage system—such as high flywheel operating speeds, a wide range of speed variations, and frequent switching of control strategies—the sliding mode surface and reaching law are redesigned. Additionally, a charge and discharge control strategy tailored for the flywheel energy storage system is developed. First, a continuous sigmoid function is established as the sliding mode switching function to accelerate system convergence and mitigate chattering. An integral sliding mode surface and a power reaching law are incorporated into the sliding mode surface function, which not only suppresses chattering but also improves the system’s dynamic response speed. Second, adaptive control of the back EMF is introduced to filter out high-frequency harmonic signals in the estimated extended back EMF, thus avoiding the chattering and phase delay caused by a low-pass filter. An improved phase-locked loop (PLL), which filters out the effects of motor rotation, is introduced to correct the impact of rotor speed variations, further enhancing the accuracy of rotor speed and position estimation. Finally, a simulation model is built on the MATLAB/Simulink platform to validate the practicality and effectiveness of the proposed control strategy.

Evaluating the fit and performance of flat-fold, cup, and three-panel respirators among Thai healthcare personnel

IntroductionHealthcare personnel (HCP) face high risks of airborne infections, including coronavirus disease 2019 (COVID-19), tuberculosis, and measles. Filtering facepiece respirators (FFRs) are critical for protection but require an adequate fit for effectiveness. Limited studies have explored the fit performance of different FFR designs in Southeast Asian populations. This study evaluates fit factors and pass rates of flat-fold, cup-shaped, and three-panel flat-fold respirators among Thai HCP and examines the influence of facial anthropometry on fit outcomes.MethodsA cross-sectional study was conducted with 223 HCP at a university hospital in Chiang Mai, Thailand. Quantitative fit testing of three NIOSH-certified N95 respirators—flat-fold, cup-shaped, and three-panel flat-fold—was performed using a TSI Portacount Pro+ 8,038 device. The Occupational Safety and Health Administration (OSHA) Condensation Nuclei Counter protocol, comprising bending, talking, and head movement exercises, was followed. Fit factors, calculated as the harmonic mean, required a passing threshold of ≥100. Twenty-two facial anthropometric dimensions were also measured. Statistical analyses included the Kruskal–Wallis test, Fisher’s exact test, and logistic regression.ResultsPass rates were 5.4% for flat-fold respirators (median fit factor [FF]: 25), 51.1% for cup-shaped models (median FF: 104), and 82.5% for three-panel flat-fold designs (median FF: 191), with significant differences (p &amp;lt; 0.001). The three-panel flat-fold maintained FF values near 200 across exercises. Anthropometric predictors varied by FFR type: head length (Adj. OR: 1.16) and nose length (Adj. OR: 1.28) influenced flat-fold models, while nasal bridge breadth (Adj. OR: 1.11) affected cup-shaped models.ConclusionThe three-panel flat-fold respirator exhibited superior adaptability, highlighting its potential as the preferred choice for Thai HCP. The low pass rate of flat-fold designs underscores the need for region-specific respirator designs. Findings emphasize the importance of localized fit testing and the development of regional fit test panels to enhance protection. Further research is needed to explore fit retention, comfort, and usability in real-world conditions.

Surf and Swash Zone Dynamics from High-Frequency Observations at a Microtidal Low-Energy Dissipative Beach

This study examines the surf and swash zone dynamics of a microtidal, low-energy, dissipative beach in Kos Island, Greece, using high-frequency optical monitoring with a Beach Optical Monitoring System (BOMS) and in situ wave measurements during the winter period. Increased wave heights induced the offshore migration of the wave-breaking zone with significant alongshore variability; however, no triggering of NOM (Net Offshore Movement) behavior was verified, while occasional rhythmic patterns were observed in the breaking location under moderate wave conditions. Shoreline dynamics showed transient erosional episodes coupled with elevated run-up excursions, yet the shoreline showed signs of recovery, suggesting a quasi-equilibrium state. Run-up energy spectra were consistently dominated by lower frequencies than those of incoming waves under both low- and high-energy conditions. This behavior is attributed to the nearshore sandbars acting as low-pass filters, dissipating high-frequency wave energy and allowing for lower-frequency motions to dominate run-up processes. A widely used empirical wave run-up predictor corresponded well with the video observations, confirming its applicability to low-energy dissipative beaches. These results underscore the role of submerged sandbars in regulating wave energy dissipation and stabilizing beach morphology under low-to-moderate wave conditions.

Built-In Error Resilient FPGA Based CIC Filter Design for Satellite Communication

Single-event and multi-bit effects are the result of radiation and ionized particles in extreme settings like space, which can lead to random failures on any electronic component. To keep the functionality of the device unaltered, these must be reduced. FPGA plays a vital role in satellite and aerospace applications in which dynamic reconfiguration essential. Cascaded Integration Comb (CIC) filters are mostly utilized in multidata signal processing and satellite communication systems as low pass filters in rate converter modules. The configuration memory of FPGA used to design CIC filter is affected with soft errors with single and multi-bit due to high radiation in higher altitude and different environment regions. The methods like triple modular redundancy (TMR) is very effective in overcoming single event transients and single-event upsets, but incur area three times of the original module. Scrubbing is a serial process method that goes over each word in memory in search of mistakes that need to be fixed. It entails a non-negligible Time to Detect (TTD) prior to repair, in which time further functionality could happen parallely and jeopardize the system. Thus, effective multi-bit error detection correction of configuration memory in FPGA is essential in maintaining the application to work for an extended time. In this research, built-in multi-bit error correction for FPGA configuration memory is proposed. The proposed work can replace time consuming scrubbing process and high area utilizing TMR for error tolerant design. To safeguard FPGA, a multi-bit error detection and correction system is performed by using multi dimensional parity with minimum area overhead. Furthermore, the suggested method can identify and rectify error when triggered by an interrupt manager reducing time to detect (TTD).

Abstract 5926: Determining classical-basal transcriptional subtypes of pancreatic ductal adenocarcinoma using circulating tumor DNA epigenomic profiling

Abstract Introduction: Classical (C) and basal (B) transcriptional subtypes are key prognostic biomarkers of pancreatic ductal adenocarcinoma (PDAC). The B subtype is associated with higher rates of metastases, chemoresistance and poor outcomes. Early data suggest that the B subtype may predict response to KRAS inhibition. Current subtyping methods rely on tissue biopsies, which may not capture subtype heterogeneity, and are difficult to implement clinically. Here, we report a novel non-invasive approach for PDAC subtyping based on profiling circulating tumor chromatin. Methods: RNA-seq, chromatin immunoprecipitation and sequencing (ChIP-seq) for H3K27ac and H3K4me3, assay for transposase-accessible chromatin sequencing (ATAC-seq), and methylated DNA immunoprecipitation sequencing (MeDIP-seq) were performed on PDAC patient derived xenografts (PDX, n = 17). C and B subtypes were called using PURIST algorithm on RNA-seq. Differential features between C and B PDXs for each epigenetic mark were identified using DESeq2. H3K27ac and H3K4me3 ChIP-seq, MeDIP-seq and low pass whole genome sequencing (LPWGS) were then performed on plasma of patients with metastatic PDAC from two clinical cohorts where C and B subtypes were previously determined. Circulating tumor DNA (ctDNA) fraction was estimated from LPWGS data using ichorCNA. A C/B score was derived for each epigenetic feature in individual patients by calculating the ratio of aggregate signal at classical:basal regions. Finally, a plasma integrated epigenomic score (PIES) was developed by combining various epigenetic features and its ability to distinguish subtypes was assessed using the area under the receiver operating characteristic (AUROC) curve. Results: Overall, 17 PDAC PDXs (C = 13, B = 4) were profiled with ChIP-seq (H3K27Ac, H3K4me3), ATAC-seq and MeDIP-seq. We identified 7, 855 differential H3K27ac sites that were also differentially accessible by ATAC-seq, 1, 603 differential H3K4me3 sites and 308 differentially methylated regions between the 2 subtypes. We then profiled 53 plasma samples from patients with metastatic PDAC who had previously undergone tissue-based RNA analysis (C = 33, B = 20). The median ctDNA fraction was comparable between the two subtypes (3.1 vs. 3.3%; p=0.4). The H3K27ac and H3K4me3 C/B scores were significantly higher in plasma from patients with classical PDAC (p&amp;lt;0.001) compared to basal PDAC. The methylation signature did not distinguish the two subtypes (p=0.2). PIES showed high discriminatory power between classical and basal samples, with an area under the curve (AUC) of 0.84 for all samples regardless of ctDNA fraction and an AUC of 0.88 for samples with &amp;gt;3% ctDNA. Conclusion: We present a non-invasive approach for distinguishing classical versus basal PDAC based on circulating tumor epigenomic profiles, which could facilitate subtyping to guide clinical decision making. Citation Format: Karl Semaan, Marc Eid, Damien Vasseur, Gunsagar S. Gulati, Ji-Heui Seo, Alexander Jordan, Andressa Dias Costa, John Canniff, Leigh Culnane, Elizabeth Andrews, Emma C. Coleman, Noa Phillips, Brad Fortunato, Hunter Savignano, Rashad Nawfal, Razane El Hajj Chehade, Talal El Zarif, Ze Zhang, Paulo Cordeiro, Antoine Hollebecque, Antoine Italiano, Brandon M. Huffman, James M. Cleary, Jonathan Nowak, Jacob E. Berchuck, Toni K. Choueiri, Kimberly Perez, Andrew J. Aguirre, Sylvan C. Baca, Brian M. Wolpin, Matthew L. Freedman, Harshabad Singh. Determining classical-basal transcriptional subtypes of pancreatic ductal adenocarcinoma using circulating tumor DNA epigenomic profiling [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 5926.

Abstract 1977: Implementation of a comprehensive clinical cfDNA analysis platform for aqueous humor liquid biopsy in retinoblastoma

Introduction: The development of liquid biopsy testing based on sequencing analysis of cell-free DNA (cfDNA) has the potential to transform cancer care in pediatric tumors, which are characterized by a wide array of genomic alterations including copy number alterations (CNAs), sequence variants, epigenetic alterations, and structural rearrangements in tumor-associated genes. In patients with retinoblastoma where direct tumor biopsy is not possible due to the risk of tumor spread, aqueous humor (AH) is a rich source of cfDNA. Comprehensive cfDNA testing, combining a low pass whole genome sequencing (LP-WGS) assay to detect CNAs and a custom targeted sequencing panel (TSP) to detect mutations, can provide a highly effective platform to inform diagnosis, risk stratification, response to therapy, and surveillance in patients with retinoblastoma. LP-WGS Results: We validated a liquid biopsy LP-WGS assay, LBSeq4Kids, to detect CNAs in cfDNA from plasma, cerebrospinal fluid, or AH. Clinical testing has been performed for the past two years for diagnostic evaluation and monitoring response to therapy using AH from retinoblastoma patients. Thirty-six children with unilateral or bilateral retinoblastoma (44 eyes) underwent LBSeq4Kids LP-WGS analysis. CNAs were detected in 30/32 (94%) samples at diagnosis or recurrence. Serial sampling of 22/44 (50%) eyes was performed using a range of 1-10 samples per patient. Seventeen of those 22 (77%) eyes demonstrated clearance of CNAs with treatment. Five eyes showed persistence of abnormal profiles, necessitating enucleation of 2 eyes. Treatment and evaluation are ongoing for the remaining three eyes. Targeted Sequencing Panel Results: We are validating a custom cfDNA TSP to detect single nucleotide variants and small indels in the coding sequence of 136 genes and fusions involving EWSR1, FOXO1, and BRAF for patients with ocular disease, solid tumors and brain tumors. To date, 17 children (21 eyes) diagnosed with retinoblastoma have had TSP testing using AH. All eleven (100%) patients with germline variants demonstrated the same RB1 alterations in the AH using the panel. Six of these patients (55%) had a second somatic alteration in RB1 detected with the panel. Two of 11 (18%) with known germline RB1 mutations also demonstrated mutations in BCOR, ARID1A, or MSH6. Somatic variants in RB1, ARID1A and BCOR were detected in the remaining six patients with sporadic retinoblastoma. Targeted sequencing of the AH surveillance samples from these patients is in progress. Conclusion: Here we demonstrate clinical utility for aqueous humor liquid biopsy molecular testing as a complementary means of diagnosis and monitoring treatment response in patients with retinoblastoma. Continued surveillance of these patients is ongoing to monitor disease status and rule out recurrence, as well as identify novel prognostic biomarkers. Citation Format: Laura A. Kagami, Eirini Christodoulou, Venkata Yellapantula, Dong Xu, Cindy Fong, Dejerianne Ostrow, Liya Xu, Jesse Berry, Jaclyn A. Biegel. Implementation of a comprehensive clinical cfDNA analysis platform for aqueous humor liquid biopsy in retinoblastoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 1977.

Cascadable Complementary SSF-Based Biquads with 8 GHz Cutoff Frequency and Very Low Power Consumption

Low-pass filters with bandwidths larger than several GHz are required in many applications, such as anti-aliasing filters in high-speed ADCs and pulse-shaping filters in high-speed DACs. In highly integrated applications, low area occupation and power consumption are key specifications, so inductor-less implementations are to be preferred. Furthermore, full CMOS implementations provide an advantage in terms of technology availability and cost. In this paper, we present an inductor-less CMOS biquad stage based on the super source follower topology that provides an 8 GHz cutoff frequency and a low power consumption of 0.42 mW per pole, showing remarkable performance also in terms of bandwidth and dynamic range. The availability of two separate current sources allows independent tuning of natural frequency and quality factor. The stage can be implemented in two complementary ways, exploiting NMOS and PMOS input devices, respectively, thus simplifying cascadability. The two complementary biquads have been implemented in the STMicroelectronics FDSOI 28 nm CMOS process and extensively simulated and provide stable performance under PVT variations and mismatches. The area occupation is about 387.5 μm2 per biquad, one of the lowest in the literature. The figures-of-merit are remarkable, as the filters achieve excellent power efficiency, very low area occupation, and good dynamic range.

The 1913 maximum temperature world record at Death Valley: Should we expect a new record in the near future?

Abstract In accordance with record high global mean temperatures new local maximum temperature records have been observed in many places worldwide over the last few years. The question arises, what is the highest 2-meter temperature on earth to be expected in the near future due to global warming? The investigation focuses on the temperature time series of Death Valley, California, the present world record holder of maximum temperature. A critical review is given on this maximum temperature world record of 134 °F or 56.7 °C, set on 10 July 1913. Different evaluations like comparison with neighbor stations, comparison with 20th century re-analyses and measures of dispersion are carried out. They all show that this record is highly questionable, because no physical mechanism is known, which could explain such a statistical outlier of the maximum temperature on a local scale over desert regions. A new low-pass filter technique of time series is being used to determine the long-term climatological temperature trend between 1911 and 2023 in Death Valley in the frame of global change, which is quite impressive. Finally, the probabilities of the occurrence of certain temperature thresholds in the near future are derived by utilizing a general extreme value distribution. It is shown that the probability of a new temperature world record of 135 °F or 57.2 °C in the next few years, which would make the present—albeit questionable—record obsolete, is still very low, despite global warming.

Enhancing Long-and Short-Term Representations for Next POI Recommendations via Frequency and Hierarchical Contrastive Learning

Next POI recommendation aids users in predicting their destinations of interest and plays an increasingly vital role in location-based social services. Recent works focus on analyzing both long-term and short-term interests in POI recommendation to gain a deeper understanding of user profiles. However, these methods for modeling long-term user’s sequences primarily rely on the Transformer model, which functions as a low-pass filter, often leading to the loss of high-frequency information. Additionally, long-term and short-term sequences are typically modeled independently, with short-term sequences often defined solely by the most recent check-ins, overlooking their interactions and dependencies. Therefore, we propose Enhancing Long-and Short-Term Representations for Next POI Recommendations via Frequency and Hierarchical Contrastive Learning (FHCRec). FHCRec captures both high-frequency and low-frequency information in long-term sequences to model richer long-term user’s preference representations. Moreover, it harnesses the characteristics of the short-term subsequences embedded within long-term sequences to enhance short-term preference characterization via local and global hierarchical contrastive learning, resulting in more personalized short-term preferences. The enhanced long-term and short-term preferences are integrated to improve model recommendation performance. Extensive experiments on three real-world datasets demonstrate the effectiveness of our method.

Deep Hypergraph Neural Networks with Tight Framelets

Hypergraphs provide a flexible framework for modeling high-order (complex) interactions among multiple entities, extending beyond traditional pairwise correlations in graph structures. However, deep hypergraph neural networks (HGNNs) often face the challenge of oversmoothing with increasing depth, similar to issues in graph neural networks (GNNs). While oversmoothing in GNNs has been extensively studied, its implications in relation to hypergraphs are less explored. This paper addresses this gap by first theoretically exploring the reasons behind oversmoothing in deep HGNNs. Our novel insights suggest that a spectral-based hypergraph convolution, equipped with both low-pass and high-pass filters, can potentially mitigate these effects. Motivated by these findings, we introduce FrameHGNN, a framework that utilizes framelet-based hypergraph convolutions integrating tight framelet transforms with both low-pass and high-pass components, as well as the commonly used strategies in designing deep GNN architecture: initial residual and identity mappings. The experiment results on diverse benchmark datasets demonstrate that FrameHGNN outperforms several state-of-the-art models, effectively reducing oversmoothing while improving predictive accuracy. Our contributions not only advance the theoretical understanding of deep hypergraph learning but also provide a practical spectral-based approach for HGNNs, emphasizing the design of multifrequency channels.

When Hypergraph Meets Heterophily: New Benchmark Datasets and Baseline

Hypergraph neural networks (HNNs) have shown promise in handling tasks characterized by high-order correlations, achieving notable success across various applications. However, there has been limited focus on heterophilic hypergraph learning (HHL), in contrast to the increasing attention given to graph neural networks designed for graphs exhibiting heterophily. This paper aims to pave the way for HHL by addressing key gaps from multiple perspectives: measurement, dataset diversity, and baseline model development. First, we introduce metrics to quantify heterophily in hypergraphs, providing a numerical basis for assessing the homophily/heterophily ratio. Second, we develop diverse benchmark datasets across various real-world scenarios, facilitating comprehensive evaluations of existing HNNs and advancing research in HHL. Additionally, as a novel baseline model, we propose HyperUFG, a framelet-based HNN integrating both low-pass and high-pass filters. Extensive experiments conducted on synthetic and benchmark datasets highlight the challenges current HNNs face with heterophilic hypergraphs, while showcasing that HyperUFG performs competitively and often outperforms many existing models in such scenarios. Overall, our study underscores the urgent need for further exploration and development in this emerging field, with the potential to inspire and guide future research in HHL.

Designing Specialized Two-Dimensional Graph Spectral Filters for Spatial-Temporal Graph Modeling

Spatial-temporal graph modeling is challenging due to the diverse node interactions across spatial and temporal dimensions. Recent studies typically adopt Graph Neural Networks (GNNs) to perform node-level aggregation at different time steps, acting as a series of low-pass graph spectral filters, for node interaction modeling. However, these filters, confined to the spatial dimension, are ill-suited for processing signals of nodes with inherent spatial-temporal interdependencies. Moreover, oversimplified low-pass filtering fails to fully exploit information from diverse node interactions. To address these issues, we propose a Spatial-Temporal Spectral Graph Neural Network (STSGNN), which designs specialized two-dimensional (2-D) graph spectral filters for comprehensive spatial-temporal graph modeling. First, based on the normalized Laplacian spectrum of spatial and temporal graphs, we extend the existing graph spectral theory from a univariate spatial dimension to a bivariate spatial-temporal dimension through a 2-D Discrete Graph Fourier Transform (2-D DGFT). Then, we leverage the bivariate Bernstein polynomial approximation, with learned basis coefficients, to design 2-D filters with specialized spectral properties for unified spatial-temporal signal filtering. Finally, the filtered signals, with refined spatial-temporal representations, are fed into well-designed pyramidal gated convolution modules to acquire multiple ranges of spatial-temporal dependencies. Experiments on traffic and meteorological prediction tasks demonstrate that STSGNN achieves state-of-the-art performance. Additionally, we visualize the 2-D filters learned from inputs with distinct spatial-temporal characteristics to enhance the model's interpretability.