next up previous
Next: Patterning [5+4* P] Up: MLB_Exercises_2008 Previous: Comparison of Learning Algorithms

Distributed computing: WTA [5 P]

Implement a Winner-Take-All mechanism with the artificial organism simulator (AOS). The task is to find and mark the input cell with the highest chemical production by placing cells of type $ Max$ at its location.

The environment consists of a 30 $ \times$ 30 grid containing a blast cell $ B_0$ and the 4 input cells $ X_1, X_2, Y_1$ , and $ Y_2$ . Input cells of type $ X_i$ and $ Y_i$ ($ i=1,2$ ) produce the chemicals $ x$ and $ y$ , respectively. Each input cell produces chemicals at a different rate. If a cell of type $ X$ ($ Y$ ) has maximum chemical production (the highest concentration of chemical $ x$ or $ y$ at its location) than output cells of type $ Max_1$ ($ Max_2$ ) should mark its location and no other output cell should be present in the environment. Therefore the type and the location of the remaining output cells indicate the type and the location of the input cell with the highest chemical production (which is the winner of the computation), respectively.

a)
Download the simulator3 and read the README.txt file. Modify the XML files config/environment_WTA.xml and config/organism_WTA.xml to program the environment and the cells.

You can use the following competences and reactors for this example:

DivideCompetence, MigrateCompetence, ConstReactor, SourceReactor, DiffSourceReactor and KillReactor.

You can visualize the results of a simulation with the MATLAB script visualize (see README.txt).

b)
Program the blast cell to divide into two types of output cells $ Max_1$ and $ Max_2$ that should search for local maxima of the chemicals $ x$ and $ y$ , respectively. Make sure that only a certain number of cells of each type is generated.

c)
Implement a mechanism that only output cells at the location of the highest chemical concentration (of the chemicals $ x$ and $ y$ ) remain after 100000 simulation time steps.

Present your results clearly, structured and legible. Document them in such a way that anybody can reproduce them effortless.


next up previous
Next: Patterning [5+4* P] Up: MLB_Exercises_2008 Previous: Comparison of Learning Algorithms
Haeusler Stefan 2009-01-19