genetic algorithms and machine learning brent harrison
TRANSCRIPT
Genetic Algorithms and Machine Learning
Brent Harrison
Genetic Algorithms Overview
• Use the concept of natural selection to optimize data.
• Initial population might not be so good…but that changes rather quickly.
Genetic Algorithm Application
• Mostly used for determining optimal parameters.
• An example, optimizing sigma values in neural nets (more on that later).
• A more fun one…optimizing theme park tours.
Traveling Salesman Problem
• A salesman must visit all cities and return to his starting location in the fastest time.
• Could try brute forcing…but seeing as there are n! permutations, this solution becomes impractical rather quickly.
Possible Answer!
• Hit it with a GA!
• Modified GA’s will produce an optimal solution most of the time for problems with up to 100,000 cities.
Machine Learning Overview
• They’re algorithms that enable machines to learn…we’ve been over this.
Types of Learning Structures
• Neural Nets:– General Regression Neural Networks– Radial Basis Function Networks– Feed Forward Neural Networks
• Naive Bayesian Classifiers
Machine Learning Applications
• Data Mining
• Breast Cancer Diagnosis
• Show how bad the BCS really is.
How Bad is the BCS?
• By using neural networks, it is possible to simulate the way that poll voters will vote.
• The predictions are based on past data freely available to anyone.
How Bad is the BCS?
• Using these simulations, we can hit the neural networks with a GA.
• By doing that, it is possible to evolve the worst BCS season possible.
• The faster we create this system…the worse the BCS is.
• Typically...within 5-10 generations we get a bad year.