Abstract:Generating learning-friendly representations for points in space is a fundamental and long-standing problem in machine learning. Recently, multi-scale encoding schemes (such as Space2Vec and NeRF) were proposed to directly encode any point in 2D or 3D Euclidean space as a high-dimensional vector, and has been successfully applied to various (geo)spatial prediction and generative tasks. However, all current 2D and 3D location encoders are designed to model point distances in Euclidean space. So when applied to large-scale real-world GPS coordinate datasets (e.g., species or satellite images taken all over the world), which require distance metric learning on the spherical surface, both types of models can fail due to the map projection distortion problem (2D) and the spherical-to-Euclidean distance approximation error (3D). To solve these problems, we propose a multi-scale location encoder called Sphere2Vec which can preserve spherical distances when encoding point coordinates on a spherical surface. We developed a unified view of distance-reserving encoding on spheres based on the Double Fourier Sphere (DFS). We also provide theoretical proof that the Sphere2Vec encoding preserves the spherical surface distance between any two points, while existing encoding schemes such as Space2Vec and NeRF do not. Experiments on 20 synthetic datasets show that Sphere2Vec can outperform all baseline models including the state-of-the-art (SOTA) 2D location encoder (i.e., Space2Vec) and 3D encoder NeRF on all these datasets with up to 30.8% error rate reduction. We then apply Sphere2Vec to three geo-aware image classification tasks -fine-grained species recognition, Flickr image recognition, and remote sensing image classification. Results on 7 real-world datasets show the superiority of Sphere2Vec over multiple 2D and 3D location encoders on all three tasks. Further analysis shows that Sphere2Vec outperforms other location encoder models, especially in the polar regions and data-sparse areas because of its nature for spherical surface distance preservation. Code and data of this work are available at https://gengchenmai.github.io/sphere2vec-website/.

VMF-SNE: Embedding for Spherical Data.

Visualizing Differences of DTI Fiber Models Using 2D Normalized Embeddings.

Multi-view Heterogeneous Fusion and Embedding for Categorical Attributes on Mixed Data.

Shape Representation and Recognition in High Dimensional Feature Space

Doubly Stochastic Neighbor Embedding on Spheres

M-Sne: Multiview Stochastic Neighbor Embedding.

Voxel2vec: A Natural Language Processing Approach to Learning Distributed Representations for Scientific Data

Voxel2vec: A Natural Language Processing Approach to Learning Distributed Representations for Scientific Data.

Sphere2Vec: A General-Purpose Location Representation Learning over a Spherical Surface for Large-Scale Geospatial Predictions

Document Classification with Spherical Word Vectors

Index $t$-SNE: Tracking Dynamics of High-Dimensional Datasets with Coherent Embeddings

ENS-t-SNE: Embedding Neighborhoods Simultaneously t-SNE

Parallel Vector Field Embedding.

Multiview Spectral Embedding

Statistical embedding: Beyond principal components

Nonparametric Spherical Regression Using Diffeomorphic Mappings

Parameter Estimation and Model-Based Clustering with Spherical Normal Distribution on the Unit Hypersphere

Spatial-Spectral Manifold Embedding of Hyperspectral Data

Use of t‐distributed stochastic neighbour embedding in vibrational spectroscopy

Clustering and Geodesic Scaling of Dissimilarities on the Spherical Surface

Efficient Algorithms for t-distributed Stochastic Neighborhood Embedding