The Data Manifold under the Microscope
Paper • 2606.15760 • Published • 2
Selected research on discovering structure in data, measuring data and representation geometry, and improving the reliability of machine-learning eval
Note Controlled benchmarks and tools for measuring the geometry of data manifolds and learned representations.
Note Structure-aware hierarchical dimensionality reduction for fast, coarse-to-fine exploration of complex datasets.
Note A position paper and benchmark examining evaluation practices and simple baselines in time-series anomaly detection.