I gave a talk to Microsoft employees on Thursday, August 13, 2015, from 1:00 to 2:00 PM. My talk was titled “Introduction to Firefly Algorithm Optimization”.
One of the coolest things about working at Microsoft is that there are an incredible number of opportunities to learn. Many people, including me, love to learn. Microsoft has daily talk sessions covering an amazing range of fascinating topics.
The talk was set up by Alex Blanton who is a manager in the Microsoft “Information Management and Machine Learning” group. They are responsible for the Microsoft Azure Machine Learning service, and also promote the dissemination of machine learning knowledge throughout the company.
One way to loosely define machine learning is “any system that uses data to make predictions”. In many cases, prediction takes the form of a mathematical equation which has constant numeric values such as 0.123 and 7.89. Determining the values of the prediction equation constants is difficult. Most traditional techniques are based on Calculus. Firefly algorithm optimization is a different approach for finding the values of a prediction equation constants.
Firefly algorithm optimization loosely mimics the behavior of fireflies. There are many simulated fireflies. The position of each firefly represents a possible solution to the problem of determining the prediction equation constants.
Firefly algorithm optimization is relatively new (the idea was first published in late 2009) and is unproven in practice. However, the technique seems to have great potential.