This chapter gives an example of a framework to solve the problem of partial observation target’sObservation target design via Machine Learning (ML)Machine learning fitting methods. More precisely, the chapter shows a general setting where the sparse nonlinear identification algorithm scLarsScLars presented in Chap.  10 is used to fit mathematical structures that solve instances of the partial state estimationState estimation problem. Two examples are used in this chapter to convey the methodology and to discuss the impact of the presence of measurement noise and parametric uncertainties on the quality of the partial estimation. The results that can be achieved using the scLarsScLars algorithm of Chap.  10 are compared to those that might be obtained using some standard MLMachine learning libraries showing the relevance of the sparse identification in terms of extrapolation capability which is intimately linked to the concept of over-fittingOver-fitting  that is fundamental to grasp in all MLMachine learning-based frameworks.