Dana Wilkinson

Papers

Total Citations

H-Index

About

Dana Wilkinson is a researcher whose work lies at the intersection of dimensionality reduction, robotics, and planning. Her most influential contribution is the introduction of **Action Respecting Embedding (ARE)** , a novel framework for dimensionality reduction that leverages action labels—information about transitions between data points—to learn low-dimensional representations. This approach is particularly powerful for sequential or time-series data, as it respects the underlying dynamics of the system. Her seminal 2005 paper on the topic has garnered **53 citations**, establishing it as a key reference in the field. Wilkinson extended this work to **subjective localization**, demonstrating how ARE can be used to build internal maps from an agent’s perspective, a crucial step for autonomous navigation. Her research also explores **learning subjective representations for planning**, where agents learn models of their environment directly from experience, reducing the need for expert-crafted models. While her citation counts are modest, the conceptual depth and originality of her work have made a lasting impact on machine learning and robotics, particularly in how agents can learn from and interact with their environments.