Search

Trajectory representation learning (TRL) maps trajectories to vectors that can then be used for various downstream tasks, including trajectory similarity computation, trajectory classification, and travel-time estimation. However, existing TRL methods often produce vectors that, when used in downstream tasks, yield insufficiently accurate results. A key reason is that they fail to utilize the comprehensive information encompassed by trajectories. We propose a self-supervised TRL framework, called RED, which effectively exploits multiple types of trajectory information. Overall, RED adopts the Transformer as the backbone model and masks the constituting paths in trajectories to train a masked autoencoder (MAE). In particular, RED considers the moving patterns of trajectories by employing a Road-aware masking strategy} that retains key paths of trajectories during masking, thereby preserving crucial information of the trajectories. RED also adopts a spatial-temporal-user joint Embedding scheme to encode comprehensive information when preparing the trajectories as model inputs. To conduct training, RED adopts Dual-objective task learning}: the Transformer encoder predicts the next segment in a trajectory, and the Transformer decoder reconstructs the entire trajectory. RED also considers the spatial-temporal correlations of trajectories by modifying the attention mechanism of the Transformer. We compare RED with 9 state-of-the-art TRL methods for 4 downstream tasks on 3 real-world datasets, finding that RED can usually improve the accuracy of the best-performing baseline by over 5%., Comment: This paper is accepted by VLDB2025

Trajectory representation learning (TRL) maps trajectories to vectors that can be used for many downstream tasks. Existing TRL methods use either grid trajectories, capturing movement in free space, or road trajectories, capturing movement in a road network, as input. We observe that the two types of trajectories are complementary, providing either region and location information or providing road structure and movement regularity. Therefore, we propose a novel multimodal TRL method, dubbed GREEN, to jointly utilize Grid and Road trajectory Expressions for Effective representatioN learning. In particular, we transform raw GPS trajectories into both grid and road trajectories and tailor two encoders to capture their respective information. To align the two encoders such that they complement each other, we adopt a contrastive loss to encourage them to produce similar embeddings for the same raw trajectory and design a mask language model (MLM) loss to use grid trajectories to help reconstruct masked road trajectories. To learn the final trajectory representation, a dual-modal interactor is used to fuse the outputs of the two encoders via cross-attention. We compare GREEN with 7 state-of-the-art TRL methods for 3 downstream tasks, finding that GREEN consistently outperforms all baselines and improves the accuracy of the best-performing baseline by an average of 15.99\%., Comment: This paper is accepted by KDD2025(August Cycle)

Accurate estimation of greenhouse gas (GHG) is essential to meet carbon neutrality targets, particularly through the calculation of direct CO2 emissions from electricity generation. This work reviews and compares emission factor-based methods for accounting direct carbon emissions from electricity generation. The emission factor approach is commonly worldwide used. Empirical comparisons are based on emission factors computed using data from the Italian electricity market. The analyses reveal significant differences in the CO2 estimates according to different methods. This, in turn, highlights the need to select an appropriate method for reliable emissions, which could support effective regulatory compliance and informed policy-making. As concerns, in particular, the market zones of the Italian electricity market, the results underscore the importance of tailoring emission factors to accurately capture regional fuel variations., Comment: Main text: 29 pages, 7 figures, 5 tables

We experimentally investigate on-chip control and analysis of spatially multimode nonlinear interactions in silicon nitride waveguide circuits. Using widely different dispersion of transverse supermodes in a strongly-coupled dual-core waveguide section, and using integrated pairs of input and output single-mode waveguides, we enable controlled excitation of nonlinear processes in multiple supermodes, while a basic physical mode decomposition aids the identification of parallel and cascaded processes. Pumping with ultrashort pulses at 1.5-$\mu$m wavelength (around 195-THz light frequency), we observe simultaneous dual-supermode, near-infrared supercontinuum generation having different spectral widths, in parallel with third-harmonic generation at around 515 nm (582 THz). Cascaded four-wave mixing with supercontinuum components upconverts the third-harmonic radiation toward a set of four shorter blue wavelengths emitted in the range between 485 and 450 nm (617 to 661 THz). The approach taken here, i.e., using chip-integrated spatial multiplexing and demultiplexing for excitation and analysis of broadband transverse nonlinear conversion, can be an advanced tool for better understanding and control in multimode nonlinear optics, such as for extending frequency conversion to wider spectral ranges via extra phase matching paths., Comment: 13 pages, 8 figures

In a predictive modular invariant theory of flavour there should exist correlations between the quark and lepton observables. So far these observables have been analyzed separately, making it impossible to investigate their interconnections. We perform for the first time a joint analysis of quark and lepton observables (22 altogether) in a modular flavour model. The model is based on $2O$ flavour symmetry and, within its class, it is characterized by the minimal number of free parameters (14 real constants). The joint analysis shows that the model is in good agreement with the experimental data for normal neutrino mass ordering, while predicting the leptonic Dirac CP-violating (CPV) phase ($\delta_{CP}$), the two Majorana CPV phases ($\eta_1$, $\eta_2$), the lightest neutrino mass ($m_1$) and the effective neutrino masses probed by beta and neutrinoless double beta decay ($m_\beta$ and $m_{\beta\beta}$). A detailed comparison of the separate (lepton-only and quark-only) and combined (lepton and quark) fit results shows differences in best-fit values and jointly allowed regions, that reflect a nontrivial interplay between quark and lepton observables in the model. Most importantly, our analysis highlights the existence of significant correlations between various pairs of such observables. For instance, the ratio of the strange and bottom quark masses, $r_{sb}$, is strongly negatively correlated with each of the three lepton mixing angles and with $\delta_{CP}$, $m_1$, $m_\beta$ and $m_{\beta\beta}$, while being positively correlated with $\eta_1$ and $\eta_2$. These findings, that are missed in separate analyses of quark and lepton flavour sectors, fall within ranges that can be tested by current and future experiments., Comment: 33 pages, 7 figures. The abstract has been slightly expanded, and the text improved in some sections for better clarity. The analysis and figures remain unchanged from the previous version

Recently, prompt learning has garnered considerable attention for its success in various Vision-Language (VL) tasks. However, existing prompt-based models are primarily focused on studying prompt generation and prompt strategies with complete modality settings, which does not accurately reflect real-world scenarios where partial modality information may be missing. In this paper, we present the first comprehensive investigation into prompt learning behavior when modalities are incomplete, revealing the high sensitivity of prompt-based models to missing modalities. To this end, we propose a novel Multi-step Adaptive Prompt Learning (MuAP) framework, aiming to generate multimodal prompts and perform multi-step prompt tuning, which adaptively learns knowledge by iteratively aligning modalities. Specifically, we generate multimodal prompts for each modality and devise prompt strategies to integrate them into the Transformer model. Subsequently, we sequentially perform prompt tuning from single-stage and alignment-stage, allowing each modality-prompt to be autonomously and adaptively learned, thereby mitigating the imbalance issue caused by only textual prompts that are learnable in previous works. Extensive experiments demonstrate the effectiveness of our MuAP and this model achieves significant improvements compared to the state-of-the-art on all benchmark datasets

In this paper, we introduce the notions of crossed module of anti-pre-Lie algebras. We explore the non-abelian cohomology of anti-pre-Lie algebras to classify their non-abelian extensions. Additionally, we investigate the inducibility of a pair of automorphisms for anti-pre-Lie algebras and construct a Wells exact sequence. Finally, we show how these results fit with abelian extensions of anti-pre-Lie algebras., Comment: 23 pages, minor corrected, continue of arXiv:2408.09971. arXiv admin note: text overlap with arXiv:2204.01060 by other authors

The modelling of temporal patterns in dynamic graphs is an important current research issue in the development of time-aware GNNs. Whether or not a specific sequence of events in a temporal graph constitutes a temporal pattern not only depends on the frequency of its occurrence. We consider whether it deviates from what is expected in a temporal graph where timestamps are randomly shuffled. While accounting for such a random baseline is important to model temporal patterns, it has mostly been ignored by current temporal graph neural networks. To address this issue we propose HYPA-DBGNN, a novel two-step approach that combines (i) the inference of anomalous sequential patterns in time series data on graphs based on a statistically principled null model, with (ii) a neural message passing approach that utilizes a higher-order De Bruijn graph whose edges capture overrepresented sequential patterns. Our method leverages hypergeometric graph ensembles to identify anomalous edges within both first- and higher-order De Bruijn graphs, which encode the temporal ordering of events. The model introduces an inductive bias that enhances model interpretability. We evaluate our approach for static node classification using benchmark datasets and a synthetic dataset that showcases its ability to incorporate the observed inductive bias regarding over- and under-represented temporal edges. We demonstrate the framework's effectiveness in detecting similar patterns within empirical datasets, resulting in superior performance compared to baseline methods in node classification tasks. To the best of our knowledge, our work is the first to introduce statistically informed GNNs that leverage temporal and causal sequence anomalies. HYPA-DBGNN represents a path for bridging the gap between statistical graph inference and neural graph representation learning, with potential applications to static GNNs.

Message passing neural networks (MPNNs) operate on graphs by exchanging information between neigbouring nodes. MPNNs have been successfully applied to various node-, edge-, and graph-level tasks in areas like molecular science, computer vision, natural language processing, and combinatorial optimization. However, most MPNNs require training on large amounts of labeled data, which can be costly and time-consuming. In this work, we explore the use of various untrained message passing layers in graph neural networks, i.e. variants of popular message passing architecture where we remove all trainable parameters that are used to transform node features in the message passing step. Focusing on link prediction, we find that untrained message passing layers can lead to competitive and even superior performance compared to fully trained MPNNs, especially in the presence of high-dimensional features. We provide a theoretical analysis of untrained message passing by relating the inner products of features implicitly produced by untrained message passing layers to path-based topological node similarity measures. As such, untrained message passing architectures can be viewed as a highly efficient and interpretable approach to link prediction.

Very-High Resolution (VHR) remote sensing imagery is increasingly accessible, but often lacks annotations for effective machine learning applications. Recent foundation models like GroundingDINO and Segment Anything (SAM) provide opportunities to automatically generate annotations. This study introduces FMARS (Foundation Model Annotations in Remote Sensing), a methodology leveraging VHR imagery and foundation models for fast and robust annotation. We focus on disaster management and provide a large-scale dataset with labels obtained from pre-event imagery over 19 disaster events, derived from the Maxar Open Data initiative. We train segmentation models on the generated labels, using Unsupervised Domain Adaptation (UDA) techniques to increase transferability to real-world scenarios. Our results demonstrate the effectiveness of leveraging foundation models to automatically annotate remote sensing data at scale, enabling robust downstream models for critical applications. Code and dataset are available at \url{https://github.com/links-ads/igarss-fmars}., Comment: Accepted at IGARSS 2024, 5 pages. Revised and corrected version

We analyze the $^{115}$In $\beta$-decay energy spectrum through the spectral moment method (SMM), previously introduced in the context of $^{113}$Cd $\beta$ decay. The spectral moments $\mu_n$ are defined as averaged $n^{\rm th}$ powers of the $\beta$ particle energy, characterizing the spectrum normalization ($n=0$) and shape ($n\geq 1$) above a given threshold. For $^{115}$In, we consider three independent datasets characterized by different thresholds. We also consider three nuclear model calculations with two free parameters: the ratio of axial-vector to vector couplings, $r=g_{\rm A}/g_{\rm V}$, and the small vector-like relativistic nuclear matrix element (NME), $s=s$-NME. By using the most recent of the three datasets, we show that the first few spectral moments can determine $(r,\, s)$ values in good agreement with those obtained by full-fledged experimental fits. We then work out the SMM results for the other datasets. We find that, although $g_{\rm A}$ quenching is generally favored, the preferred quenching factors may differ considerably depending on the chosen experimental data and nuclear models. We discuss various issues affecting both the overall normalization and the low-energy behaviour of the measured and computed spectra, and their joint effects on the experimentally quoted half-life values. Further $^{115}$In $\beta$-decay data at the lowest possible energy threshold appear to be crucial to clarify these issues., Comment: 14 pages, including 8 figures

Discrete charge, parity, and time symmetries (C, P, and T) of quantized fermion states are extended by a triality symmetry (t), producing the CPTt Group, transforming between three generations of fermions., Comment: 17 pages, 4 figures

Commonsense question answering has demonstrated considerable potential across various applications like assistants and social robots. Although fully fine-tuned pre-trained Language Models(LM) have achieved remarkable performance in commonsense reasoning, their tendency to excessively prioritize textual information hampers the precise transfer of structural knowledge and undermines interpretability. Some studies have explored combining LMs with Knowledge Graphs(KGs) by coarsely fusing the two modalities to perform Graph Neural Network(GNN)-based reasoning that lacks a profound interaction between heterogeneous modalities. In this paper, we propose a novel Graph-based Structure-Aware Prompt Learning Model for commonsense reasoning, named G-SAP, aiming to maintain a balance between heterogeneous knowledge and enhance the cross-modal interaction within the LM+GNNs model. In particular, an evidence graph is constructed by integrating multiple knowledge sources, i.e. ConceptNet, Wikipedia, and Cambridge Dictionary to boost the performance. Afterward, a structure-aware frozen PLM is employed to fully incorporate the structured and textual information from the evidence graph, where the generation of prompts is driven by graph entities and relations. Finally, a heterogeneous message-passing reasoning module is used to facilitate deep interaction of knowledge between the LM and graph-based networks. Empirical validation, conducted through extensive experiments on three benchmark datasets, demonstrates the notable performance of the proposed model. The results reveal a significant advancement over the existing models, especially, with 6.12% improvement over the SoTA LM+GNNs model on the OpenbookQA dataset.

Recommendation systems play a crucial role in various domains, suggesting items based on user behavior.However, the lack of transparency in presenting recommendations can lead to user confusion. In this paper, we introduce Data-level Recommendation Explanation (DRE), a non-intrusive explanation framework for black-box recommendation models.Different from existing methods, DRE does not require any intermediary representations of the recommendation model or latent alignment training, mitigating potential performance issues.We propose a data-level alignment method, leveraging large language models to reason relationships between user data and recommended items.Additionally, we address the challenge of enriching the details of the explanation by introducing target-aware user preference distillation, utilizing item reviews. Experimental results on benchmark datasets demonstrate the effectiveness of the DRE in providing accurate and user-centric explanations, enhancing user engagement with recommended item., Comment: 5 pages, 2 figures

Consider a finite primitive solvable group. We observe that a result of Y. Yang implies that there exist two points whose pointwise stabilizer has derived length at most $9$. We show that, if the group has odd cardinality, then there exist two points whose pointwise stabilizer is abelian., Comment: 6 pages, Added reference to Yang's paper

We present experimental results highlighting two key differences resulting from the choice of training algorithm for two-layer neural networks. The spectral bias of neural networks is well known, while the spectral bias dependence on the choice of training algorithm is less studied. Our experiments demonstrate that an adaptive random Fourier features algorithm (ARFF) can yield a spectral bias closer to zero compared to the stochastic gradient descent optimizer (SGD). Additionally, we train two identically structured classifiers, employing SGD and ARFF, to the same accuracy levels and empirically assess their robustness against adversarial noise attacks., Comment: 6 Pages, 4 Figures; Accepted in the International Conference on Scientific Computing and Machine Learning