3 papers
cs.LG2025
Modular Jets for Supervised Pipelines: Diagnosing Mirage vs Identifiability
Suman Sanyal
Classical supervised learning evaluates models primarily via predictive risk on hold-out data. Such evaluations quantify how well a function behaves on a distribution, but they do…
cs.LG2025
Perception Learning: A Formal Separation of Sensory Representation Learning from Decision Learning
Suman Sanyal
We introduce Perception Learning (PeL), a paradigm that optimizes an agent's sensory interface using task-agnostic signals, decoupled from downstre…
cs.LG2025
Convexity-Driven Projection for Point Cloud Dimensionality Reduction
Suman Sanyal
We propose Convexity-Driven Projection (CDP), a boundary-free linear method for dimensionality reduction of point clouds that targets preserving detour-induced local non-convexity.…