Empirically Measuring Transfer Distance for System Design and Operation
This work addresses data scarcity issues for practitioners in fields such as machine prognostics and defense, offering a systems engineering approach to enhance transfer learning, though it appears incremental in its methodological contribution.
The paper tackles the challenge of limited data in transfer learning for domains like machine prognostics and defense by introducing a formal definition of transfer distance to empirically quantify model transferability, demonstrating its application in designing machine rebuild procedures and predicting operational performance in computer vision.
Classical machine learning approaches are sensitive to non-stationarity. Transfer learning can address non-stationarity by sharing knowledge from one system to another, however, in areas like machine prognostics and defense, data is fundamentally limited. Therefore, transfer learning algorithms have little, if any, examples from which to learn. Herein, we suggest that these constraints on algorithmic learning can be addressed by systems engineering. We formally define transfer distance in general terms and demonstrate its use in empirically quantifying the transferability of models. We consider the use of transfer distance in the design of machine rebuild procedures to allow for transferable prognostic models. We also consider the use of transfer distance in predicting operational performance in computer vision. Practitioners can use the presented methodology to design and operate systems with consideration for the learning theoretic challenges faced by component learning systems.