over 4 years ago by Loke Zhi Ming
Hi Sir, thanks for the lesson, really appreciate it. Wanted to ask somewhere around 10:00 in the video, you mentioned something about squaring the numbers so that it "punishes" when the number reaches too high. May I ask for a further explanation regarding that one and how it affects the solver ? As understand that you have already set the fitness value to be 0 before running Galapagos. So how does Squaring the numbers affect the solver ? Would really appreciate the help, thanks
Arie Willem de Jongh over 4 years ago
Hi Loke,
Excellent question! So ideally the value coming out is 0, this is the optimum. So let's take a list of values from good to bad:
The difference between each of them is only 1, so when the solver gets an output of 4 and then the next iteration it's 3 it improved the score but only by 1.
To make this more dramatic we square those outcomes so instead, we get:
Now when there is an iteration and let's say the result is 25 and then the next iteration it's 16. The improvement is 9 (25-16) instead of only 1 (5-4). So for the solver this more dramatic increase in the score makes it go to the optimum more aggressive.
Best,
Arie
Ning about 4 years ago
I was wondering about it too, thanks Arie!