On the efficiency of machine learning for fatigue assessment of post-processed additively manufactured AlSi10Mg