Codetown ::: a software developer's community
Time: March 24, 2016 from 6pm to 8pm
Location: PowerDMS @ Church St Station
Street: PowerDMS, 101 S Garland Ave #300
City/Town: Orlando, FL 32801
Website or Map: http://www.orlandojug.com
Event Type: meeting
Organized By: Michael Levin
Latest Activity: Mar 21, 2016
This month we have Dave Moskowitz in from Sarasota for a fascinating talk.
Here's the abstract for the presentation. There is a possibility of massive amounts of pizza!
Genetic programming (GP) is a class of nature-inspired AI algorithms that aims to automatically generate a population of computer programs to solve a target problem where the solution is unknown. By defining a fitness function (a high-level statement of the problem) and applying principles of Darwinian evolution (mating, survival of the fittest), GP breeds a population of computer programs that achieve a higher level of fitness and come closer to an optimal solution over time.
GP is most applicable to problems without deterministic solution. An example of such a problem is stock market prediction. There is no canonical theory on how to "beat the market". The most widely accepted belief regarding this question is stated in the efficient market hypothesis (EMH), which holds that no method can consistently beat the overall market, However, there is conflicting research from both the economic and AI communities regarding the validity of this hypothesis. While most research appears to deny the possibility of consistently beating the market after factoring in transaction costs, recent no transaction cost trading platforms, such as Robinhood, as well as more powerful and available data processing capabilities, make beating the market more in reach than ever.
The presentation will begin with an introduction to genetic programming, focusing on the more common tree-based (LISP) representation. Other approaches to GP, as well differences between genetic programming and genetic algorithms will also be discussed.
A Java based GP system, developed by the speaker, will be used to illustrate some basic GP examples. Finally, the genetic programming algorithm will be run on the S&P 500 index using current market conditions to see what it predicts for the immediate future.
This talk will convince you that genetic programming is essentially an approachable technique that can be applied towards many problems in the area of numeric optimization, prediction, design, or just about anything else you can do in a computer program.
About the Speaker:
David Moskowitz is a Market Data Architect at Ned Davis Research in Venice Florida where he develops and maintains web and REST based financial data delivery systems. He is currently a PhD candidate in Computer Science at Nova Southeastern University and is also the former chairman of the Sarasota Java Users Group.
Feel free to invite a friend. Please RSVP so I'll know how much pizza to order.