I want to see something beautiful, so I tried to visualize the function used for benchmarking the optimization function.
reference
https://en.wikipedia.org/wiki/Test_functions_for_optimization
Execution source
https://github.com/akidukin/optimize_test/blob/master/benchmark_functions.py
Target
--Not decided (self-satisfaction) ――People who think "Hey, there is something like this"
What is the function used for benchmarking optimization functions?
What is an optimization function? https://ja.wikipedia.org/wiki/最適化問題
As a rough image
- x + 10 = 25
- x + 60 = 15
- x + 45 = 60 A function that makes the machine find x accurately There are some indicators about the performance of the optimization function and the hogehoge for measuring the performance of the above function, and it seems that speed, accuracy, robustness, etc. are difficult.
Simply put, measuring performance with a simple (one-dimensional and single solution) object like the one above does not change, so it is easier to compare performance with a more complex object. ..
What did you do
I just created it with the function scratch on the reference URL and executed it At that time, it is designed to accept only input up to 2 dimensions.
Result (Function name is not official. Correctly go to reference URL)
Impressions
I like the shape of the RosenBrock function for some reason. (Konami) Next, let's look at the trajectory of finding the optimum solution with some typical optimization functions for each.