Defensive Signal Processing: The Case for the Use of Nonparametric and Robust Statistical Methods to Reduce Product Liability Exposure [post]

Michael Lang
2017 unpublished
This paper makes the case that in an Internet of Things (IoT) world where data processing hasbecome pervasive, the assessment of whether or not the underlying (statistical) modeling assumptions are justified and appropriate should no longer be limited to the perspective of mathematical statistics alone. The paper argues that large parts of sound academic research in engineering lack practical merit in that, akin to a concept car, they are not market-ready, most commonly due to feasibility and
more » ... o feasibility and liability issues. Through an analysis of both statistical and legal aspects it will be shown that the stoic pursuit of ‘optimality’ more often than not yields to risky and suboptimal outcomes when applied to actual physical world problems. To address this, the concept of ‘Defensive Signal Processing’ is introduced and future research directions are brieflyoutlined.
doi:10.20944/preprints201709.0054.v1 fatcat:iixxy642lzhs7mvsmcwkzqrodq