MIND the Gap：一种具有可调空间尺度的地理隐式神经表示
MIND the Gap: A Geographic Implicit Neural Representation with Adjustable Spatial Scale
- Taylor Geospatial（泰勒地理空间）
- University of British Columbia（不列颠哥伦比亚大学）
- University of Colorado Boulder（科罗拉多大学博尔德分校）
- University College London（伦敦大学学院）
- Vector Institute（向量研究所）
- Technical University of Munich（慕尼黑工业大学）
- University of Bonn（波恩大学）
- University of Texas at Austin（德克萨斯大学奥斯汀分校）
- Washington University in Saint Louis（圣路易斯华盛顿大学）
- Arizona State University（亚利桑那州立大学）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
针对地理稀疏测量数据，提出MIND方法，通过马特罗什卡隐式神经蒸馏将专家模型嵌入蒸馏为可调空间粒度的通用坐标嵌入，并引入CoordBench评估套件，在区域保留预测中取得最优性能。
AI中文摘要：
地理测量数据往往稀疏，导致我们想要绘制的量在大片区域缺乏标签。地理隐式神经表示（INR）通过学习平滑、通用的嵌入来解决这一问题，这些嵌入可以在任意坐标处查询。下游模型将这些嵌入与稀疏标签相结合，在推理时无需卫星图像即可预测未采样位置的目标值。然而，尽管对于遥感应用至关重要，向遥远区域的泛化在很大程度上仍未得到探索。我们引入了马特罗什卡隐式神经蒸馏（MIND），该方法将来自专家预训练地理模型的嵌入蒸馏为具有可调空间粒度的单一通用坐标嵌入。MIND在多个嵌入维度上使用嵌套监督，这些维度定义了一系列连续块。在我们的实验中，早期块捕获较粗略的地理变化，而后期块添加更精细的细节。下游预测器可以仅保留前导块，或者使用我们的分块惩罚（Chunked Penalty）进行拟合，以在保留完整嵌入的同时降低后期块的权重，而无需重新训练INR。为了衡量MIND并与全球现有方法进行比较，我们引入了CoordBench，这是一个大规模INR评估套件，包含52个数据集和78个目标，旨在测试各种空间尺度下局部插值和保留区域的预测。MIND及其分块惩罚变体在测试的INR中取得了最高的聚合回归和分类分数，并在区域保留下取得了总体最高分数，为地理INR树立了新的最先进水平。
英文摘要：
Geographic measurements are often sparse, leaving large areas without labels for the quantities we want to map. Geographic implicit neural representations (INRs) provide coordinate-based embeddings that can be combined with sparse labels to predict at unsampled locations without satellite imagery at inference. Yet existing INRs are largely evaluated with random holdouts, leaving their ability to generalize across larger geographic gaps unclear. We introduce Matryoshka Implicit Neural Distillation (MIND), a geographic INR whose spatial granularity can be adjusted after training. MIND distills several pretrained geospatial models using nested supervision at increasing embedding dimensions, dividing the representation into contiguous chunks. Early chunks capture broad spatial patterns, while later chunks add increasingly local variation. Downstream models can retain only the leading chunks or use our Chunked Penalty to reduce reliance on later chunks without retraining the INR. We also introduce CoordBench, comprising $52$ datasets and $78$ targets with both random and regional holdouts at multiple spatial scales. Across CoordBench, fine-scale features help most when labels are nearby, while smoother representations generalize better across larger geographic gaps. MIND with the Chunked Penalty achieves the highest aggregate regression and classification performance among tested INRs and the highest overall performance under regional holdout. These results show that geographic representations should be evaluated and adapted according to the spatial separation between labeled and prediction locations.
