Researchers present PPDL, a framework for predicting channel-level user retention ratios in multi-channel paid user acquisition. The method integrates physical priors with deep learning to capture the sharp early churn typical of retention curves. Accurate early forecasts are intended to help marketers allocate advertising budgets more effectively.
A new arXiv paper examines whether deep learning and explainable AI methods can help diagnose hepatocellular carcinoma and pinpoint biomarkers across five stages of disease progression. The work uses a transcriptomic biomarker dataset for liver cancer, aiming for models whose predictions can be traced to interpretable biological features. The authors frame it as an early exploration of combining accuracy with explainability in cancer diagnostics.
A new arXiv paper introduces LEAPSC, a method that combines deep joint source-channel coding with in-loop concept erasure to prevent learned semantic features from leaking sensitive attributes like gender, race, or speaker identity. The approach aims to preserve transmission efficiency while removing private information during encoding. It is a research contribution rather than a released product.
A new arXiv paper proposes a credit risk early warning system that merges heterogeneous data sources using deep learning and real-time analytics. The authors argue that existing financial monitoring tools are slowed by fragmented data and delayed detection. The work is categorized under machine learning and is cross-listed on arXiv.
A new arXiv preprint studies whether multi-task learning can improve predictive process monitoring, which forecasts how ongoing organizational processes will unfold so information systems can support proactive analysis rather than just execution. The work situates itself within deep learning approaches that have already raised prediction accuracy in this area, and explores the potential of training related prediction tasks jointly.
Researchers present a semantic segmentation deep learning framework for large-scale bioacoustic detection, applied to fin whale calls recorded by ocean-bottom seismometers. The approach repurposes seismic instruments, originally deployed for geophysics, as a broad-area passive acoustic monitoring network. This could expand monitoring of baleen whales over long time spans and wide ocean regions.
A new arXiv preprint describes a deep learning pipeline that uses 3D gait recordings to help identify autism spectrum disorder, aiming to sidestep the subjectivity and cost of standard clinical assessments. The approach adds attention mechanisms to the model and evaluates results across multiple data folds to test how stable the performance is. The work is presented as a step toward more objective, non-invasive screening tools, though the abstract only covers the motivation and method rather than deployment.
Researchers present a hybrid temporal convolutional network and transformer architecture designed to forecast satellite collision probability early, using conjunction data messages as input within a dedicated analysis framework. The work targets the growing operational load on satellite operators caused by more frequent close approaches in low Earth orbit as satellites and debris multiply.
A new arXiv preprint argues that treating all input noise as equivalent may limit the robustness of neural networks. The work focuses on geophysical and seismic data, where heterogeneous noise from active field sites can hide weak events and hinder automated analysis. It suggests that robustness evaluations should account for differences in noise characteristics rather than assuming a uniform perturbation model.
A new arXiv survey examines deep learning methods for cleaning noise from gravitational-wave detector data, arguing that techniques must cope with the full range of spinning and precessing binary systems. The paper notes that matched filtering remains the established approach but comes with trade-offs that learned models aim to address. Reconstructed waveforms feed into parameter estimation, tests of general relativity and population studies.
A new arXiv paper evaluates deep learning-based channel estimation for 5G NR links to low Earth orbit satellites. The authors note that while such methods work well in terrestrial networks, satellite links add Doppler shifts and synchronization problems that likely demand network-specific designs. The work presents a comparative evaluation of HELENA against other approaches in this non-terrestrial setting.
A revised arXiv paper presents a deep evidential regression approach for estimating forest height using multimodal satellite imagery. The method targets applications including carbon accounting, biodiversity monitoring, and ecosystem management. It aims to give accurate predictions while quantifying uncertainty in sparse-data settings.
A new arXiv preprint argues that parameter counts and Hessian rank are insufficient for measuring the effective complexity of deep neural networks, since many different parameter settings produce identical predictions. The work analyzes the symmetries and singular structure of the loss landscape to better characterize model complexity. It appears in the cs.LG category as a new submission.
A new arXiv paper argues that deep learning research on automated emotion recognition from electrocardiogram signals is hard to compare because studies differ in preprocessing, training and evaluation setups. The authors present a unified evaluation framework intended to allow fairer, more direct comparison between architectures. The work is a cross-listing on arXiv's machine learning section.
A new arXiv paper introduces WaVeFuse, a hybrid deep learning approach for forecasting stock market indices. The method targets three issues in existing models: noise from OHLCV data leaking into derived technical indicators, treating all channels the same during multi-scale decomposition, and mismatched frequency signals. It applies channel-wise wavelet denoising with vertical attention fusion to adapt across market regimes.
A new preprint introduces a deep learning approach for tracking heart muscle motion in ultrasound images that embeds physiological constraints into the model. The design is intended to reduce temporal drift across the cardiac cycle, improving the reliability of strain measurements used to assess cardiac function. The work appears on arXiv as paper 2609.09577.
Researchers have introduced PSCT-Net, a neural method that builds 3D skull CT volumes from sparse bi-planar X-rays, offering a lower-dose option for diagnosing craniofacial conditions in children. The approach pairs a differentiable back-projection module with attention-guided refinement to make the highly ill-posed reconstruction more geometry-aware.
Researchers present an uncertainty-aware workflow that fuses geophysical inversion results with drilling data via Bayesian deep learning to model copper mineralization in three dimensions. Applied to the Kogodai prospect in the Rudny Altai region, the method quantifies prediction uncertainty to help prioritize exploration drilling in structurally complex terrain where sampling is sparse.
Researchers propose a multi-pass, multi-view blended learning approach for generating full 3D chest CT volumes from a single 2D chest radiograph. The task is an ill-posed inverse problem, made harder by the limited availability of paired X-ray and CT training data. The paper builds on earlier methods that relied on digitally reconstructed radiographs to overcome data scarcity.
Researchers introduce DFNN, a deep neural network framework designed for regression tasks where the response variable lies in a general metric space rather than Euclidean coordinates. The approach targets non-Euclidean outputs such as probability distributions, networks, symmetric positive-definite matrices, and compositional data. It extends Fréchet regression concepts into deep learning architectures for these increasingly common data types.
A new arXiv preprint describes a hardware-aware deep learning method for spotting electrical faults and power quality disturbances in More Electric Aircraft. The authors note that most existing diagnostics were built for conventional 50/60 Hz grids and may not transfer to the high-frequency networks used on aircraft. The work aims to support faster, more reliable monitoring of these onboard systems.
A new arXiv paper proposes analyzing how deep neural networks learn by studying them at the level of functions rather than individual parameters. The authors borrow tools from statistical mechanics, treating parameter configurations as microscopic states to explain why networks with billions of weights show consistent, predictable learning patterns. The approach aims to provide a theoretical framework for understanding training dynamics in very large models.
An arXiv paper explores using deep learning to interpret finger-specific movement intentions from surface electromyography (sEMG) signals in stroke survivors. Because measurable muscle activity often persists even when movement is weak or incomplete, these signals can serve as control inputs for rehabilitation hardware. The study frames the problem as five-finger multilabel intent decoding.
Researchers present a combined deep learning and finite element framework that factors in the real cross-sectional geometry of extruded concrete filaments when evaluating whether printed layers can bear the weight of subsequent ones. The work addresses a common limitation of existing buildability assessments, which typically rely on simplified filament shapes, potentially misjudging the stability of 3D-printed concrete structures.
A revised preprint examines how the modular structure of relational graphs interacts with the depth of neural networks when learning from graph-structured data. The author situates the work within graph-based machine learning, including graph neural networks and reinforcement learning, and analyzes how graph structure shapes learning outcomes. The version posted is an update to an earlier draft.
A revised arXiv paper in machine learning introduces Relation, a token-mixing mechanism that splits an operation standard attention fuses into a single score-to-flow step. The method first organizes pairwise evidence into explicit Self and Exchange relations and then allocates information flow across them. The authors position this as an alternative to dominant attention-based token mixing in sequence models.
A new overview article on arXiv surveys the theory of overparameterized machine learning, in which models with far more parameters than training examples still achieve strong performance. The paper explains how such behavior conflicts with the classical bias-variance tradeoff and organizes recent theoretical work developed to explain it. It serves as a structured introduction for readers interested in the statistical foundations of modern deep learning.
Researchers have introduced PRIME-SVR, a physics-informed implicit reconstruction method that turns motion-corrupted 2D fetal MRI slices into high-resolution 3D brain volumes with quantitative T2 mapping. The approach builds on standard slice-to-volume reconstruction by adding multi-echo signal modeling, targeting improved quantitative imaging of the developing fetal brain.
A new arXiv paper examines how divergence-based regularization relates to Sharpness-Aware Minimization, two widely used techniques for improving generalization in deep learning. The authors use local curvature analysis to explain why f-divergence regularization, like SAM, improves robustness to parameter perturbations.
A new arXiv paper presents Muon-C, a variant of the Muon optimizer designed specifically for convolutional layers. The authors argue that applying Muon through standard matrix unfolding misrepresents the geometry of convolution, since it describes a local patch mapping rather than the convolution operator itself. Their approach instead computes the orthogonalized update direction in a way that aligns with the operator's true structure.
A new arXiv preprint presents a deep learning approach for spotting cyberattacks in Low-Earth Orbit satellite communication networks, which face constantly shifting and complex conditions. The method combines temporal and multimodal modeling, departing from standard network intrusion detection techniques built for more static terrestrial environments. The work is a cross-listed submission and has not yet been peer reviewed.
A new arXiv paper introduces GRIPNet, a detection architecture that incorporates a Gaussian radial intensity prior to help locate pulmonary nodules in CT scans. The authors target lesions smaller than six millimeters, which remain difficult for existing detectors despite being key to early lung cancer diagnosis through low-dose CT screening. The work is a research contribution rather than a released product.