## Abstract

This is a proposal for mathematically determining the learning rate to be used in a deep supervised convolutional neural network (CNN), based on student fluency. The CNN model shall be tasked to imitate how students play the game “Packet Attack”, a form of gamification of information security awareness training, and learn in the same rate as the students did. The student fluency shall be represented by a mathematical function constructed using natural cubic spline interpolation, and its derivative shall serve as the learning rate for the CNN model. If proven right, the results will imply a more human-like rate of learning by machines.