Are we going to do a profiling/time report?

[leopoldoagorio]

We had discussed this but i don't remember if we reached a veredict

[bbazzano]

I'd say yes for the bi-material model. A minimal proposal could be:

• For the MLP model, measure training time, average prediction time over a number of examples.
• For the onsas FEA analysis, an average analysis time over the generation of the training datapoints.

With those bits of information, we can determine how many predictions have to be made until the MLP route starts to be faster than Onsas FEA. And in the long run view, how much faster MLP is at predicting than Onsas FEA.

Let me know your thoughts!

[leopoldoagorio]

This goes to future work de facto