In this presentation, we will try to answer the following question: How can we compute the best possible approximation to a given multivariate function when some of its additional characteristic properties as convexity are known? We also discuss some ongoing related research.