arXiv:2607.04949v1 Announce Type: new Abstract: We study the problem of k-means clustering on large datasets. The state-of-the-art for the problem is given by coresets-based approaches, which build small weighted summaries of the input and derive approximate solutions with rigorous quality guarantees from them. One of the most popular and advanced approaches to derive coresets for k-means is sensitivity sampling. However, sensitivity sampling requires to compute the importance of each input point with respect to the whole dataset over all possible choices of centers. Since the exact computation of such quantities is unfeasible, current approaches work by approximating the sensitivity values. Nevertheless, the runtime of such approaches is still impractical for large datasets. In this work, we propose to reduce the runtime of sensitivity-based approaches for k-means by leveraging predictions to approximate the importance of input points. We first formally prove that current theoretical results on coresets construction via sensitivity sampling hold for coarser approximations of sensitivities compared to the one required by existing approaches. This implies that even fairly noisy predictors can be leveraged for sensitivity-sampling approaches. We then propose a natural predictor, which applies to the common scenario where clustering is performed (over time) on a sequence of datasets from the same problem. We prove that when the datasets in the sequence come from the same (unknown) distribution, centers resulting in a low error on one dataset can be used as predictions for sensitivity sampling in subsequent datasets, with guarantees on their quality. We perform an extensive experimental evaluation showing that our approach significantly improves, in terms of clustering cost vs runtime, over uniform sampling and state-of-the-art sensitivity sampling approaches when applied to sequences of datasets.
Science Journals
arXiv:2607.04457v1 Announce Type: new Abstract: Reconstructing the pulmonary vascular tree from computed tomography (CT) images is essential for quantitative lung analysis, vascular morphology assessment, and patient-specific modeling, yet it remains challenging because vessels span multiple scales, from proximal arteries to distal microvasculature. Clinical chest CT is further affected by limited spatial resolution, partial volume effects, heterogeneous image quality, and respiratory motion artifacts. Unlike deep learning-based pulmonary vessel segmentation methods that require large annotated datasets, we propose a deterministic, training-free, and explainable pipeline for CT-based pulmonary vascular tree reconstruction. The method fuses multiscale Hessian-based Frangi and Sato vesselness filters using a weighted maximum response across 12 spatial scales from 1 to 8 mm, enabling detection of large pulmonary arteries and peripheral branches. Lung parenchyma is segmented by Hounsfield unit thresholding, morphological post-processing, and Chan-Vese active contour refinement. Vascular centerlines are extracted using the Kimimaro implementation of the TEASAR algorithm; separate left- and right-lung vascular graphs are then constructed, pruned, and verified for acyclicity. Geometric plausibility is assessed using volumetric fractal dimension, Strahler order analysis, Horton ratios, and Murray's law. The resulting fractal dimension of approximately 2.3 is consistent with reported values for the human pulmonary vasculature. At the same time, residual deviations in branching metrics reflect distal-vessel truncation caused by finite CT resolution. These results indicate that the proposed explainable pipeline can generate geometrically plausible pulmonary vascular tree models and may support quantitative pulmonary imaging, vascular morphometry, and computational lung modeling.
arXiv:2607.04459v1 Announce Type: new Abstract: We extend the distributional framework developed in the companion paper [Zolla, arXiv:2605.15716] to the nonlinear case, focusing on the second-order ($\chi^{(2)}$) response responsible for second-harmonic generation (SHG). Starting from the most general tensorial nonlocal second-order constitutive relation and combining a spatial moment expansion with a distributional thin-layer limit, we show that the full complexity of the nonlinear interfacial response condenses, at leading order, into two scalars, the nonlinear surface susceptibilities $\chi^{(2),s}_\parallel$ and $\chi^{(2),s}_\perp$, associated with the tangential and normal components of the electric field, respectively. A key structural result is established: via a marginal integration over one field argument, the nonlinear surface problem reduces recursively to an effective linear one, whose surface susceptibility is determined by the bulk nonlinear kernel alone. Generalized nonlinear Maxwell boundary conditions are derived explicitly for planar and spherical interfaces, and curvature corrections are obtained systematically. The formalism is illustrated on Gaussian, Yukawa, and tensorial Lorentz kernels.
arXiv:2607.04461v1 Announce Type: new Abstract: Inference-time scaling for text-to-image generation has progressed from simple Best-of-$N$ (BoN) sampling to guided search methods that verify and steer candidate trajectories at intermediate denoising steps. These approaches focus on when and how often to verify during denoising but largely treat the cost of generation itself as fixed. Moreover, the standard practice of comparing methods by number of function evaluations (NFEs) counts only denoising forward passes and ignores verifier overhead, which can distort efficiency rankings. We show that under wall-clock evaluation, simple BoN already matches or outperforms several guided search techniques, suggesting that compute is better spent on broader exploration than on repeated intermediate verification. This motivates Flash-BoN, which generates a large pool of inexpensive draft candidates by combining three complementary acceleration knobs: timestep truncation, layer skipping, and activation proxies into a single configuration optimized once per model. An efficient multi-stage verification procedure then identifies the most promising draft, which is refined at full quality. Across three benchmarks and three model scales, Flash-BoN consistently outperforms all baselines under fixed wall-clock budgets, with gains that grow at larger model scales (+8% AUC). We further show that our strategy combines well and improves existing orthogonal techniques such as reflection-based prompt optimization (+16% AUC). The gains correlate with increased candidate diversity, which also enables draft-guided selection to accelerate RL post-training convergence.
arXiv:2607.04463v1 Announce Type: new Abstract: The performance of acoustic machine learning systems is commonly evaluated using fully annotated test sets. In real-world deployments, however, exhaustively labeling large volumes of continuously collected audio data is often infeasible. Consequently, performance assessment typically relies on a small labeled subset of the available data, introducing a sampling bias that can severely distort evaluation metrics. This paper studies methods for compensating the bias in evaluation-labeled subsets under strict annotation-budget constraints. We study whether importance weighting techniques can mitigate this discrepancy by compensating for the selection bias. Specifically, we implement and compare three density-ratio estimation methods: kernel density estimation (KDE), logistic regression, and k-nearest neighbors (kNN), utilizing feature-space representations of the deployed audio. To emulate realistic deployment scenarios, the labeled subsets are generated using five distinct sampling strategies based on active learning techniques. Experiments conducted on an audio scene classification (ASC) benchmark demonstrate that importance weighting consistently yields more realistic accuracy estimates, significantly reducing the gap between subset-based metrics and the true evaluation performance.
arXiv:2607.04824v1 Announce Type: new Abstract: We study a sequential learning problem for stable matchings in two-sided markets where preferences on both sides are initially unknown. We focus on a centralized setting where an algorithm matches agents at each time step and receives noisy rewards that reflect the preferences of the matched agents, following a semi-bandit feedback structure. We adopt a pure exploration perspective, aiming to efficiently identify the optimal stable matching with high probability. Our work extends prior results by handling \emph{two-sided uncertainty} and by exploiting \emph{partial preference} information. A central ingredient is the notion of \textbf{pervasive stable matching}, which enables the identification of optimal stable matchings under partial preferences. We propose elimination-based algorithms whose stopping criteria exploit the structure of the learned partial preferences, and provide a refined sample-complexity analysis. Beyond pure exploration, we extend our approach to regret minimization and establish regret bounds with respect to the \emph{optimal} stable matching that avoid dependence on the minimum reward gap $\Delta_{\min}$.
arXiv:2607.04827v1 Announce Type: new Abstract: Biological theory usually treats time as an external chronological variable against which growth, aging, and ecological change are parametrized. Yet living systems also generate an internal measure of duration through physiological cycling and irreversible entropy production, and the regularities of allometric lifespan scaling, biological clocks, life-history evolution, ecological synchronization, and disease are ordinarily studied in isolation rather than within a single thermodynamic internal-time framework. The Principle of Biological Time Equivalence (PBTE) proposes such a framework.
arXiv:2607.04829v1 Announce Type: new Abstract: The integration of Non Terrestrial Networks into 5G and beyond cellular systems has introduced a significant paradigm shift, enabling ubiquitous connectivity and extending services to previously unconnected and underserved remote regions. In particular, Low Earth Orbit satellites, operating close to the Earth surface, can provide communication latency comparable to that of terrestrial networks. However, due to their high mobility, LEO satellites trigger frequent handovers, which degrade users quality of experience and increase signaling overhead. In this work, our objective is to minimize the number of handovers in a LEO satellite system while preventing satellite overloading. We formulate the problem within a game theoretic framework and apply the Spatial Adaptive Play algorithm to obtain a handover efficient and load balanced solution. Additionally, we propose a low complexity heuristic algorithm to achieve similar objectives with reduced computational overhead.
arXiv:2607.04832v1 Announce Type: new Abstract: In recent years, studies have revealed a decline in semantic variety across popular music lyrics, particularly in English-language songs on streaming platforms like Spotify. This research examines whether a similar trend can be observed in a different linguistic and cultural context: the lyrics of all finalist songs from the 75 editions of the Sanremo Music Festival, Italy's most renowned music competition. What sets this work apart is the development of a flexible and efficient methodology for tracking changes in semantic similarity over time, which can be applied to different datasets to study similar phenomena. Drawing on a combination of full-text, segment-based, topic-based, and word-level analyses, the approach leverages both embedding techniques and large language models. When applied to the Sanremo corpus, this framework reveals a gradual move toward increasing semantic uniformity, echoing the global patterns identified in previous studies. These findings underscore the value of natural language processing tools in uncovering long-term shifts in musical language and cultural expression.
arXiv:2607.04512v1 Announce Type: new Abstract: We consider compactly supported RBFs having algebraically decaying Fourier transforms. Here we focus especially on generalized Wendland RBFs and their modification by making them smoother away from zero, a process we call eccentric smoothing. Specifically, we consider mapping properties of the integral operators (sometimes known as covariance operator) for these novel type RBFs. Moreover, we show that eccentric smoothing makes wavelet-inspired compression technique for the kernel matrix feasible.
arXiv:2607.04834v1 Announce Type: new Abstract: While complex-pole hierarchical equations of motion (HEOM) have dramatically expanded the reach of numerically exact quantum dynamics simulations of open quantum systems, they suffer from numerical instabilities rooted in the non-Hermitian structure of their Liouvillian. Yet, the origin of this structure remains obscure. Here, we report a previously unknown gauge freedom in complex-pole HEOM: a continuous family of analytically equivalent Liouvillians, all encoding the same bath correlation function, whose numerical properties vary dramatically. This gauge controls both the eigenspectrum and non-normality of the hierarchy generator, revealing spectral divergence and non-normal error amplification as two distinct instability mechanisms. By optimizing this gauge, we introduce GO--HEOM, which eliminates divergences in strongly coupled Brownian oscillator environments and extends numerically exact simulations of sub-Ohmic dynamics -- including through the delocalized-to-localized quantum phase transition -- to previously inaccessible coupling strengths. Because this gauge transformation is independent of the bath-correlation decomposition scheme, our GO--HEOM becomes a general, broadly compatible strategy for accessing numerically exact quantum dynamics of open quantum systems over arbitrary coupling and highly non-Markovian regimes.
arXiv:2601.20248v2 Announce Type: replace Abstract: One of the fundamental ways of representing a droplet shape is through its Rayleigh-modes, where each mode corresponds to distinct surface-energy. Previous studies have focused on the effect of these modes on free oscillations of drops. In this paper, we systematically quantify how the different prescribed initial axisymmetric Rayleigh modes modulate aerodynamic energy uptake and the resulting deformation of an impulsively accelerated drop. Using experimentally validated VOF-based multiphase numerical simulations, we isolate the coupled effects of finite-amplitude surface oscillation modes and the associated initial surface-energy state by initializing the drops with well-defined $(n,0)$ modes and phases $\{0,\pi\}$, while conserving the equivalent drop volume. We find that the deformation outcome is governed by the drag due to the drop's initial geometry, and the dynamic coupling between the free modal oscillations and the forced aerodynamic deformation. We find that constructive superposition amplify deformation, whereas destructive superposition can stabilize the drop even when the aerodynamic forcing is sufficient to deform an analogous spherical drop to breakup. Initial modes and phases that channel a larger fraction of the input power into deformation, in the form of oscillatory kinetic energy and additional surface energy, attain larger deformations and are closer to the fragmentation threshold. These coupling effects are especially pronounced in high-viscosity systems, where viscous dissipation is large and facilitates the transfer of a larger fraction of the total energy to translational kinetic energy instead of oscillatory kinetic energy. For low density-ratio systems, early-time coupling and energy transfer is the dominant mechanism that governs drop deformation.
arXiv:2602.08842v3 Announce Type: replace Abstract: As highly automated driving is transitioning from single-vehicle closed-access testing to commercial deployments of public ride-hailing in selected areas (e.g., Waymo), automated driving and connected cooperative intelligent transport systems (C-ITS) remain active fields of research. Even though simulation is omnipresent in the development and validation life cycle of automated and connected driving technology, the complex nature of public road traffic and software that masters it still requires real-world integration and testing with actual vehicles. Dedicated vehicles for research and development allow testing and validation of software and hardware components under real-world conditions early on. They also enable collecting and publishing real-world datasets that let others conduct research without vehicle access, and support early demonstration of futuristic use cases. In this paper, we present karl., our new research vehicle for automated and connected driving. Apart from major corporations, few institutions worldwide have access to their own L4-capable research vehicles, restricting their ability to carry out independent research. This paper aims to help bridge that gap by sharing the reasoning, design choices, and technical details that went into making karl. a flexible and powerful platform for research, engineering, and validation in the context of automated and connected driving. More impressions of karl. are available at https://karl.ac.
arXiv:2602.24266v2 Announce Type: replace Abstract: Which internal mechanisms of a neural network can be replaced while preserving the computation it performs? Structured pruning asks for smaller deployable networks; causal abstraction asks for high-level models that commute with interventions. We introduce causal mechanism reduction (CMR), a framework that treats a trained network as a deterministic structural causal model and replaces selected internal variables by constants or affine functions of retained variables. These replacements compile exactly into smaller dense networks by bias and weight folding, and induce reduced causal models testable with interchange interventions. We derive a unified second-order replacement-risk objective whose special cases recover mean replacement, variance-based pruning (VBP), logit-distortion scoring, and affine neuron merging, together with a margin-based certificate linking logit distortion to interchange-intervention agreement. The framework also exposes a basic invariance requirement: functionally identical ReLU networks should induce the same reduction. Under exact positive-scaling reparameterizations, VBP's kept set collapses to chance-level overlap while the logit-distortion score is exactly invariant. Empirically, CMR variants are competitive with VBP under matched fine-tuning of DeiT-Tiny on ImageNet-100; the clearer separation appears in the invariance and interchange tests, where the logit-distortion score preserves kept sets and consistently improves distributional fidelity. CMR thus gives pruning, compilation, and causal-abstraction verification a common object to optimize and verify.
arXiv:2603.23115v2 Announce Type: replace Abstract: The realism of AI-generated images (AIGI) poses increasing challenges for reliable forensic detection, where heterogeneous expert detectors may produce conflicting predictions across diverse generative sources and post-processing conditions. Existing multi-expert fusion methods rely on fixed rules or learned fusion strategies, offering limited ability to assess sample-specific reliability, execute rigorous adjudication of conflicts, and provide evidence-grounded explanations. We propose AgentFoX, an LLM-driven agentic multi-expert framework for AIGI detection that employs a command-and-reasoning core to perform evidence fusion. Following predefined guidelines, the core coordinates designated subtasks to collect semantic and signal-level evidence, reason over structured contexts to determine authenticity, and generate an auditable report for explainability. During this process, Expert Profiles are constructed for model-centric reliability assessment, while Clustering Profiles are built for data-centric contextual analysis, jointly establishing evidence contexts for conflict resolution. Extensive evaluations across diverse benchmarks demonstrate the robustness and generalizability of AgentFoX under complex conditions.
arXiv:2511.12853v2 Announce Type: replace-cross Abstract: Brain tumors induce complex structural deformations that obscure the patient' s original neuroanatomy, making it difficult to distinguish tumor-induced changes from inherent anatomical variability. Reconstructing a subject-specific pseudo-healthy brain can provide a critical reference for such analysis, but this task is inherently counterfactual, as paired pre-tumor scans and explicit healthy guidance are unavailable. We propose BrainNormalizer, a diffusion-based framework for subject-specific pseudo-healthy brain MRI reconstruction that enables anatomy-informed reconstruction without requiring paired data or explicit healthy references. The framework learns anatomical priors and edge-based structural conditioning through a two-stage training strategy consisting of inpainting-based diffusion fine-tuning and ControlNet-based edge conditioning. At inference, counterfactual pseudo-healthy reconstruction is achieved through a deliberate misalignment strategy, where tumorous inputs are paired with non-tumorous prompts and mirrored contralateral edge maps. This allows subject-specific anatomical guidance to be constructed from the patient's own anatomy, enabling anatomically consistent pseudo-healthy reconstruction that preserves individual structural characteristics. Experiments on the BraTS2020 dataset demonstrate that BrainNormalizer achieves improved distributional realism, symmetry-based structural consistency, and reduced false positive detection compared to existing methods. These results indicate that the proposed framework provides a principled approach for subject-specific counterfactual reconstruction and supports downstream analysis of tumor-induced deformation.
arXiv:2512.11415v3 Announce Type: replace-cross Abstract: We show that nonequilibrium dynamics can play a constructive role in unsupervised machine learning by inducing the spontaneous emergence of latent-state cycles. We introduce a model in which visible and hidden variables interact through two independently parametrized transition matrices, defining a Markov chain whose steady state is intrinsically out of equilibrium. Likelihood maximization drives this system toward nonequilibrium steady states with finite entropy production, reduced self-transition probabilities, and persistent probability currents in the latent space. These cycles are not imposed by the architecture but arise from training, and models that develop them reproduce the empirical distribution of data classes more faithfully, with a clear correlation between agreement with the data and entropy production. Compared with equilibrium approaches such as restricted Boltzmann machines, our model breaks the detailed balance between the forward and backward conditional transitions and relies on a log-likelihood gradient that depends explicitly on the last two steps of the Markov chain. Hence, this exploration of the interface between nonequilibrium statistical physics and modern machine learning suggests that introducing irreversibility into latent-variable models can improve the fidelity of the generated data distribution.
arXiv:2512.17342v4 Announce Type: replace-cross Abstract: We initiate the study of nowhere-zero flow reconfiguration. The natural question is whether any two nowhere-zero $k$-flows of a given graph $G$ are connected by a sequence of nowhere-zero $k$-flows of $G$, such that any two consecutive flows in the sequence differ only on a cycle of $G$. We study this problem in the setting of integer flows and group flows, and prove a number of positive and negative results. * The natural reconfiguration variant of Tutte's 5-flow conjecture, stating that any two nowhere-zero 5-flows in any 2-edge-connected graph are connected, is false in the group and integer cases. * All nowhere-zero $\mathbb{Z}_2^8$-flows of every 2-edge-connected graph are connected and for every sufficiently large abelian group $A$, all nowhere-zero $A$-flows of every 2-edge-connected graph are connected. * The group structure affects the answer, contrary to the existence problem for nowhere-zero flows. * We highlight a duality with recoloring in planar graphs and deduce that any two nowhere-zero 7-flows in a planar graph are connected, among other results. * For every 2-edge-connected graph $G$, there is an integer $k$ such that all nowhere-zero $k$-flows of $G$ are connected.
arXiv:2604.08942v3 Announce Type: replace-cross Abstract: Budding-cell packings jam before all buds are mechanically constrained, so the post-jamming state is set not by the pressure $P$ alone but also by the fraction $u$ of buds that remain unconstrained. We develop a mean-field theory in these two variables for this regime. Stress feedback suppresses growth on the loaded buds, so continued growth is redirected onto the remaining free ones. As those buds are completed they add contacts that raise the excess coordination without a comparable rise in prestress. We introduce a modified Maxwell count, a bud-depletion relation, and a flux-partition argument to predict the post-jamming coordination, the density at which the reservoir of initially free buds is exhausted, and how strong feedback can stiffen the packing while generating little internal pressure. Because the added contacts raise the rigidity while the prestress stays low, we conclude that feedback-regulated growth provides a distinct mechanism of self-rigidification.
arXiv:2607.04540v1 Announce Type: new Abstract: Geometry-conditioned 3D scene generation enables the creation of 3D environments from user-provided geometry, offering direct control over scene structure and object layout. To generate such 3D scenes, current methods commonly adopt a three-stage design that first defines a view schedule, then synthesizes multi-view observations along the scheduled views, and finally reconstructs a 3D representation from the generated images. However, defining the view schedule becomes a major bottleneck for outdoor scenes, where large, unstructured, and unbounded geometry makes it difficult to obtain views that provide sufficient coverage while supporting stable generation. To address this bottleneck, we present SceneFrom3D, a framework that automatically schedules views from outdoor input geometries. SceneFrom3D constructs a directed generation graph whose nodes represent anchor views and whose edges represent interpolation trajectories, defining which views to synthesize, which view pairs to interpolate, and in which order generation should proceed. Beyond automatic view scheduling, SceneFrom3D further improves controllability through object-level conditioning, assigning each object an identity image for appearance guidance and a geometry-adherence parameter for region-wise control over the input geometry. Experiments demonstrate that SceneFrom3D achieves state-of-the-art geometry-conditioned outdoor 3D scene generation, producing high-quality scenes with controllable object appearance and geometry adherence.
arXiv:2607.04541v1 Announce Type: new Abstract: Camera-radar (CR) fusion is a practical sensing configuration for autonomous driving, but existing models are typically trained with task-specific supervision, limiting reusable representation learning. We present CRISP, a spatiotemporal CR backbone pretrained through forecasting-based representation learning. Given historical multi-view images and radar sweeps, CRISP learns a unified bird's-eye-view (BEV) representation by predicting future LiDAR point clouds. LiDAR is used only as privileged supervision during pretraining; the deployed model requires only camera and radar. To make forecasting-based pretraining effective for CR fusion, CRISP introduces an enhanced radar encoder, radar-enhanced temporal self-attention, and multimodal feature rendering with modality innovation gating. These components inject radar range and Doppler cues into BEV temporal propagation and allow BEV tokens to selectively incorporate camera and radar evidence. Experiments on nuScenes show that CRISP improves long-horizon point cloud forecasting and transfers effectively to downstream tasks, including 3D detection, tracking, online mapping, motion forecasting, future occupancy prediction, and planning, suggesting that predictive CR pretraining is a promising path toward scalable driving representations under practical sensor configurations. The project website is https://umfieldrobotics.github.io/CRISP.
arXiv:2506.18801v2 Announce Type: replace Abstract: Subway rail vibrations are a major source of structural deterioration, environmental noise, and passenger discomfort in urban railway systems. Here, we present a broadband design methodology for railway tuned mass dampers (TMDs) based on the concept of joining multiple locally resonant bandgaps. The proposed framework begins with an experimental modal analysis of a UIC60/60E1 rail to identify its dominant vibration modes and develop an equivalent dynamic model. Guided by the proposed bandgap-joining strategy, single- and multi-resonator TMDs are subsequently designed, fabricated, and experimentally validated before being implemented on the railway track. The experimental investigation demonstrates that the tuned single-resonator configuration reduces the vibration amplitudes at the dominant resonances by up to 74\%, while incorporating a second resonator further broadens the effective attenuation bandwidth, confirming the advantages of the proposed multi-resonator concept. Finally, a random vibration analysis is performed to evaluate the effectiveness of the designed TMDs under stochastic excitations representative of practical railway operating conditions, predicting an average reduction of approximately 12\% in the RMS vibration response. The proposed methodology provides a practical framework for translating locally resonant metamaterial concepts into compact, manufacturable, and non-invasive railway vibration absorbers with enhanced broadband vibration mitigation capabilities.
arXiv:2506.19136v4 Announce Type: replace Abstract: We show that the out-of-equilibrium driving protocol of score-based generative models (SGMs) can be learned via local learning rules. The gradient with respect to the parameters of the driving protocol is computed directly from force measurements or from observed system dynamics. As a demonstration, we implement an SGM in a network of driven, nonlinear, overdamped oscillators coupled to a thermal bath. We first apply it to the problem of sampling from a mixture of two Gaussians in 2D. Finally, we train an oscillator network on the MNIST dataset to generate images of handwritten digits 0 and 1.
arXiv:2506.23845v2 Announce Type: replace Abstract: While sparse autoencoders (SAEs) have generated significant excitement, a series of negative results have added to skepticism about their usefulness. Here, we establish a conceptual distinction that reconciles competing narratives surrounding SAEs. We argue that even if SAEs may be less effective for \textit{acting on known concepts}, SAEs are especially powerful tools for \textit{discovering unknown concepts}. This distinction separates existing negative results from positive results, and suggests several classes of SAE applications. Specifically, we outline use cases for SAEs in (i) ML interpretability, explainability, fairness, auditing, and safety, and (ii) social and health sciences.
TransitNet: A Compact Attention-Augmented Deep Learning Framework for Low-SNR Transit Blind Searches
arXiv:2606.18932v2 Announce Type: replace-cross Abstract: Motivated by the observational incompleteness of intermediate-to-long-period Earth-size planets, we present TransitNet, a compact attention-augmented deep-learning framework for low-SNR transit blind searches. To enable realistic method development and objective threshold calibration under blind-search conditions, we develop a unified dataset construction, benchmarking, and threshold-selection framework. On recovery benchmarks constructed from unseen Kepler targets, TransitNet attains 95.2 percent accuracy in the challenging SNR range of 6 to 8 and outperforms both TLS and BLS, achieving ROC-AUC and PR-AP values of 0.974 and 0.982, respectively. In an injected Earth-size and sub-Earth-size transit recovery experiment, TransitNet achieves a recovery rate of 93.0 percent, substantially exceeding those of TLS (63.1 percent) and BLS (60.0 percent). In addition to detection, TransitNet provides attention-based estimates of transit windows and midpoints. On an independent evaluation set, 97.4 percent of injected transits are fully covered by the estimated transit window. Applied to real Kepler observations, the model successfully recovers all 34 selected confirmed Kepler planets, with a mean absolute transit midpoint error of 1.24 hours. The model combines a compact footprint of about 1.5 MB with high inference efficiency, yielding speed-ups of about 12 to 25 times relative to CPU-TLS and about 4 to 5 times relative to CPU-BLS. These results demonstrate that TransitNet provides an accurate, scalable, and computationally efficient framework for low-SNR transit blind searches in the tested regime and motivate its extension to longer-period Earth-size planet searches.