Wondering if this approach could I found myself be applied to CFD (computational flow dynamics) methods to reduce the total volume of data points while still getting approximately close to the correct final answer?