Selected work

Representative Papers

Earth Observation Foundation Models for Terrestrial Ecohydrology: From Representation Learning to Process Inference

Aug 15, 2026

This study addresses scale mismatches and insufficient inference credibility in Earth observation foundation models for ecohydrology by proposing a process-aware framework that integrates multi-source remote sensing with process-based models to establish an observation-inference hierarchy. Through meta-analysis and benchmark auditing to identify model gaps, the research employs hybrid workflows and label-efficient adaptation techniques for multi-scale representation learning. The findings delineate model applicability boundaries, revealing pre-training data biases and critical spectral band deficiencies. Furthermore, the work validates the effectiveness of spatial context and hybrid processing pipelines, significantly enhancing both the credibility and interpretability of monitoring coupled water-energy-carbon dynamics.

0 citationsRead paper

PhenoEmbed: Self-Supervised Multispectral UAV Time-Series Embeddings for Individual Tree Crown Phenology

Jul 11, 2026

This study addresses the challenge of robustly characterizing tree crown phenology, which is hindered by dynamic spectral, textural, and boundary changes throughout the growing season that confound conventional AI models. To overcome this, the work proposes the first application of self-supervised temporal embedding for individual tree crown phenology modeling, leveraging multi-temporal UAV multispectral imagery. The approach integrates contrastive learning, masked image reconstruction, and explicit seasonal time encoding to produce compact 256-dimensional embeddings. By aligning and cropping crown polygons followed by principal component analysis, the model is trained on 5,885 samples from the HeideBench dataset. The resulting embeddings capture meaningful phenological variation, with the first two principal components explaining 25.1% of the total variance and achieving a median top-1 cosine similarity of 0.946 in nearest-neighbor retrieval—significantly outperforming handcrafted features and mean-pooling baselines.

0 citationsRead paper

TiledAttention: a CUDA Tile SDPA Kernel for PyTorch

Mar 02, 2026

This work proposes TiledAttention, the first schedulable scaled dot-product attention (SDPA) forward operator built on cuTile Python (TileIR), which bridges the gap between high performance and kernel-level customization flexibility in PyTorch. TiledAttention enables direct manipulation of tiling shapes, staging strategies, and shared memory layouts at the Python level without requiring modifications to CUDA or CUTLASS templates. By integrating online softmax computation, streaming K/V tile loading, and NVIDIA’s CUDA Tile technology, it is seamlessly exposed through a PyTorch-callable interface. Experimental results on DGX GB10 demonstrate that TiledAttention consistently outperforms the standard eager execution path across diverse sequence lengths, head dimensions, and FP16/BF16 precisions, delivering both high performance and reproducibility—making it well-suited for real-world research and deployment scenarios.

0 citationsRead paper

Criteria-first, semantics-later: reproducible structure discovery in image-based sciences

Feb 17, 2026

This work addresses the vulnerability of existing image analysis methods to label drift under semantic-priority paradigms, which compromises their reliability in open science, cross-sensor/cross-site comparability, and long-term monitoring. To overcome this limitation, the authors propose a novel “standard-first, semantics-later” paradigm that decouples structural discovery from semantic mapping. By leveraging cybernetics, the principle that observation entails distinction, and information theory, they formulate explicit optimization criteria to extract stable, semantics-agnostic structures—such as partitions, structural fields, or hierarchies—prior to aligning them with domain-specific ontologies. This framework ensures that structural outputs remain independent of labeling schemes, enabling multiple interpretations and long-term interoperability. Validated across diverse domains, the approach demonstrates broad applicability in scenarios where labels are non-scalable, thereby advancing structural findings as FAIR, AI-ready digital objects suitable for digital twins and continuous monitoring.

0 citationsRead paper

DeepKoopFormer: A Koopman Enhanced Transformer Based Architecture for Time Series Forecasting

Aug 04, 2025

To address the poor interpretability and instability under noise of Transformer models in time-series forecasting for high-dimensional nonlinear systems, this paper proposes DeepKoopFormer—a novel architecture integrating Koopman operator theory with the Transformer framework. Its encoder-propagator-decoder structure models system dynamics as a spectrally constrained linear Koopman operator in a latent space. To ensure stability and interpretability, it incorporates three key mechanisms: a bounded spectral radius constraint, Lyapunov-energy-based regularization, and orthogonal parameterization of the Koopman operator. Experiments on synthetic dynamical systems and real-world multivariate time series—including climate, financial, and power-grid data—demonstrate that DeepKoopFormer consistently outperforms LSTM and standard Transformers. Notably, it achieves substantial gains in both noise robustness and long-horizon forecasting accuracy, while providing physically meaningful, linearized spectral representations of underlying nonlinear dynamics.

0 citationsRead paper

Recent publications

Latest Papers

Earth Observation Foundation Models for Terrestrial Ecohydrology: From Representation Learning to Process Inference

Aug 15, 2026

This study addresses scale mismatches and insufficient inference credibility in Earth observation foundation models for ecohydrology by proposing a process-aware framework that integrates multi-source remote sensing with process-based models to establish an observation-inference hierarchy. Through meta-analysis and benchmark auditing to identify model gaps, the research employs hybrid workflows and label-efficient adaptation techniques for multi-scale representation learning. The findings delineate model applicability boundaries, revealing pre-training data biases and critical spectral band deficiencies. Furthermore, the work validates the effectiveness of spatial context and hybrid processing pipelines, significantly enhancing both the credibility and interpretability of monitoring coupled water-energy-carbon dynamics.

0 citationsRead paper

PhenoEmbed: Self-Supervised Multispectral UAV Time-Series Embeddings for Individual Tree Crown Phenology

Jul 11, 2026

This study addresses the challenge of robustly characterizing tree crown phenology, which is hindered by dynamic spectral, textural, and boundary changes throughout the growing season that confound conventional AI models. To overcome this, the work proposes the first application of self-supervised temporal embedding for individual tree crown phenology modeling, leveraging multi-temporal UAV multispectral imagery. The approach integrates contrastive learning, masked image reconstruction, and explicit seasonal time encoding to produce compact 256-dimensional embeddings. By aligning and cropping crown polygons followed by principal component analysis, the model is trained on 5,885 samples from the HeideBench dataset. The resulting embeddings capture meaningful phenological variation, with the first two principal components explaining 25.1% of the total variance and achieving a median top-1 cosine similarity of 0.946 in nearest-neighbor retrieval—significantly outperforming handcrafted features and mean-pooling baselines.

0 citationsRead paper

TiledAttention: a CUDA Tile SDPA Kernel for PyTorch

Mar 02, 2026

This work proposes TiledAttention, the first schedulable scaled dot-product attention (SDPA) forward operator built on cuTile Python (TileIR), which bridges the gap between high performance and kernel-level customization flexibility in PyTorch. TiledAttention enables direct manipulation of tiling shapes, staging strategies, and shared memory layouts at the Python level without requiring modifications to CUDA or CUTLASS templates. By integrating online softmax computation, streaming K/V tile loading, and NVIDIA’s CUDA Tile technology, it is seamlessly exposed through a PyTorch-callable interface. Experimental results on DGX GB10 demonstrate that TiledAttention consistently outperforms the standard eager execution path across diverse sequence lengths, head dimensions, and FP16/BF16 precisions, delivering both high performance and reproducibility—making it well-suited for real-world research and deployment scenarios.

0 citationsRead paper

Criteria-first, semantics-later: reproducible structure discovery in image-based sciences

Feb 17, 2026

This work addresses the vulnerability of existing image analysis methods to label drift under semantic-priority paradigms, which compromises their reliability in open science, cross-sensor/cross-site comparability, and long-term monitoring. To overcome this limitation, the authors propose a novel “standard-first, semantics-later” paradigm that decouples structural discovery from semantic mapping. By leveraging cybernetics, the principle that observation entails distinction, and information theory, they formulate explicit optimization criteria to extract stable, semantics-agnostic structures—such as partitions, structural fields, or hierarchies—prior to aligning them with domain-specific ontologies. This framework ensures that structural outputs remain independent of labeling schemes, enabling multiple interpretations and long-term interoperability. Validated across diverse domains, the approach demonstrates broad applicability in scenarios where labels are non-scalable, thereby advancing structural findings as FAIR, AI-ready digital objects suitable for digital twins and continuous monitoring.

0 citationsRead paper

DeepKoopFormer: A Koopman Enhanced Transformer Based Architecture for Time Series Forecasting

Aug 04, 2025

To address the poor interpretability and instability under noise of Transformer models in time-series forecasting for high-dimensional nonlinear systems, this paper proposes DeepKoopFormer—a novel architecture integrating Koopman operator theory with the Transformer framework. Its encoder-propagator-decoder structure models system dynamics as a spectrally constrained linear Koopman operator in a latent space. To ensure stability and interpretability, it incorporates three key mechanisms: a bounded spectral radius constraint, Lyapunov-energy-based regularization, and orthogonal parameterization of the Koopman operator. Experiments on synthetic dynamical systems and real-world multivariate time series—including climate, financial, and power-grid data—demonstrate that DeepKoopFormer consistently outperforms LSTM and standard Transformers. Notably, it achieves substantial gains in both noise robustness and long-horizon forecasting accuracy, while providing physically meaningful, linearized spectral representations of underlying nonlinear dynamics.

0 citationsRead paper