Least-squares, Sentinels and Substractive Optimally Localized Average

D edi e a Jacques-Louis Lions,avec toute mon amiti e, pour son soixante dixi eme anniversaire, en souvenir des ann ees INRIA. Abstract. We present with uniied notations three approaches to linear parameter estimation: least-squares, sentinels, and Substrative Optimally Localized Average (SOLA). It becomes then obvious that the two last approaches correspond to the very same mathematical problem. This a new interpretation to sentinels, brings new computational tools to SOLA, and makes clear their link to the classical least-squares approach.