Learning Complex Classifier Systems: The Very Beginning. What’s this all about?

Author:

February 10, 2012

Our world is made up of many different things, of many different magnitudes that interact amongst themselves with a resultant outcome. These interacting things themselves may yet be composed of other smaller things. Confused yet? Indeed this is the description of a complex system. The complexity can become even more complex by considering dimensions of time, location and even the presence or lack thereof rules dictating these interrelations. Our world is undeniably quite complex; the endeavor to model it in order to predict consumer behavior, create customer segmentation or analyze trends (to name a few) is an extremely daunting task.

Why bother then? Simple; solution focused problem solving!

To solve problems, we need to first model it. To solve my problem, it needs to be modeled around my world. Me, me, me! In short, personalization is key.

This begs the question of how to go about personalizing the modeling of my problem when modeling the world in general is a daunting task. The response to this will be to focus on the rules that describe your world from a microscopic view. For example, rather than model my mornings in some highly complicated statistical model, I would use semantic descriptors. Semantic descriptors could look like, ‘wake up’, ‘meditate’ (I usually need a rest from multi-dimensional thinking), ’take shower’, ‘drink coffee’. Moreover, adding other aspects such as time and location in my descriptors increase the complexity as aforementioned. Thus by paying attention to the microscopic views of our system, and letting them interact under rules that are specific to me, a pattern emerges over time that defines my world.

If you ever watched ‘The Matrix’ movies, and wondered how Neo could go from a 0-1 mathematical matrix representation to being able to suspend mid-air, you can now attribute that as an emergent behavior from his learning complex system represented by the Matrix.

Stay tuned next week, when we dive even deeper into adaptive behavior models and how they can help you engage your customers.

About Mechie Nkengla

Mechie Nkengla Ph.D. is a Senior Algorithm Scientist at Talk3. She has extensive research experience in Mathematics and Computer Science with focus on massive data. Specifically her interest includes, solving multidimensional data-intensive problems that arise in autonomous, predictive and dynamic complex models of systems in applied industries. In particular, she relishes designing mathematical algorithms and formulating principled and efficient solutions to automate such analysis. Oh, and yes, she also enjoys long walks on the beach and dancing during warmer weathers.
Share |

One Response to “Learning Complex Classifier Systems: The Very Beginning. What’s this all about?”

  1. Norbert C

    very interesting. A very simplistic way of explaining the complexity of a “system” . At least, so far. Looking forward to the next bit.

    Reply

Leave a Reply