What is the metric you used to calculate the “similarity of high level feature charts” in your paper? Thank you for your help.