Question 1: 

I used many different combinations on all three testing formats, I found out that the ones with the more neurons ran the fastest but also had double or even triple amount the neurons of the effiecent answer I got, to find this out I used a table method like 2*2 or 3*3*3 gave me the fastest times on averages. I also tried changing the nuerons locations, but that gave me roughly the same times. I tested 10 times for each case, I had 4 bad ones, 2 Medium, 2 excellent, and 1 perfect or the best I got. Here is my data: https://docs.google.com/spreadsheets/d/10ObfBTQ8pQUDNtaB8dk2gIo8eXLd_Kjby9i7Vx0TDuY/edit?gid=0#gid=0 

Question 2:

Yes, I did train multiple to get many different results. For all 3 I had tested many test cases of course the fastest would always be maxed out on layers and neurons. I had found many effiencient lower neuron examples with really fast speeds only using half of the neurons. But I had one superior testing model that destryod the answer for all of these, that had 4 layer 5 neurons in each layer.

​