2. Evolve To Solve | TSA STATES 2015
Genetic Algorithm Basics
Genetic Algorithms have been used in a wide range of fields, from
bioinformatics to economics to physics.
Usually non-reusable, used in very specific circumstances for specific
problems.
Complex, requiring understanding of complicated programming practices
Can be very slow and computer intensive
Can solve complex, hard to understand problems
Basic form of computer learning - The computer learns how to solve the
problem.
What does this all mean in regards to our project?
4. What makes Evolve to Solve Better?
● Mobility
○ Evolve to Solve allows the use of Genetic Algorithms for any purpose the
user wishes without requiring tedious rewrites. The framework is also
small and lightweight, making it great for even memory-sensitive projects.
Size of framework, post build: 11 kb
● Simplicity
○ Evolve to Solve is easy to implement, requiring only a limited knowledge of
computer science, allowing students or those without much experience to
explore the world of computer learning. While most genetic algorithms are
slow and cumbersome, Evolve to Solve is quick and utilizes threading to
prevent locking up
Evolve To Solve | TSA STATES 2015
6. Proof of Concept
As a proof of concept, our team used Evolve To Solve to
try to generate copies of images using random changes. While less
applicable to a real world problem, this example proves the
concept of Evolve To Solve valid, even for problems with
thousands of possible permutations.
10
Original
Image:
ETS 80%
completion
ETS Solution
(takes about 10 seconds)
8. Evolve to Solve: Solving complex problems with computer learning
Computer Learned Solutions
Each
was generated in
only a few minutes.
Scores range from
900 - 1,200 points.
Humans scored
600 - 700 points
All
solutions were self-
taught by the
algorithm without
the need for excess
problem solving
code.
Evolve To Solve | TSA STATES 2015