As simple as a makefile

$ cd CPPN-NEAT
$ make clean 
$ make

We've incorporated both of the minimal actions related to evolving a network, i.e. training a certain network & using it.

In order to train a certain network ( in this example we use genetic_encodings/default_00.genetic_encoding ) the -t flag must be used followed by the file name in which the network is stored (in JSON format).

$ cd CPPN-NEAT 
$ ./executables/CPPN-NEAT -t genetic_encodings/default_00.genetic_encoding

This will create a file in the same folder with the same name ( with extension .out.genetic_encoding ) which contain the final network structure, i.e. the champion.

* We use JSON as a network structure in order to make a visual dashboard in the future using 'html' and 'jvascript'.

You may also use the -p flag in order to print in the console the champion's fitness and the champion's network structure.

$ cd CPPN-NEAT 
$ ./executables/CPPN-NEAT -t genetic_encodings/default_00.genetic_encoding -p

Once the network has been trained we can use it by specifying the -x flag followed by the file name ( in this example we use genetic_encodings/default_00.out.genetic_encoding ) in which the network is stored (in JSON format).

The last flag must be the value's specification flag -v followed by the number of inputs needed by the network (an error will arrise if the number of inputs differ from the number of given values)

$ cd CPPN-NEAT 
$ ./executables/CPPN-NEAT -x genetic_encodings/default_00.out.genetic_encoding -v 0 0

This will return in the console the respective outputs of the network.

The network structure is extremately simple. Again we will be using the genetic_encodings/default_00.genetic_encoding file.

To see more information about JSON go to the following link

For the node definition we have:

"nodes":
		[
			{"exist": 1,"node": 0,"type": 0, "row": 0, "function": "IDENTITY"},
			{"exist": 1,"node": 1,"type": 0, "row": 0, "function": "IDENTITY"},
			{"exist": 1,"node": 2,"type": 2, "row": 1, "function": "SIGMOID"}
		]

Where:

• exist indicates if the node exists (1) or if it doesn't (0). This is used for the algoithm in order to make the mating esear. For a starting network all the nodes are considered to exist.
• node represent the node's id. They must be correlative but not necesary ordered.
• type indicates the node type, i.e. 0: INPUT, 1: HIDDEN & 2: OUTUT
• row indicated the row in which the node is positioned. This is used for the algorithm in order to avoid recurrent connections. If two nodes are connected they must be in different rows.
• function indicates the node's function. There are (so far) a set of functions, i.e. IDENTITY, SIGMOID, GAUSSIAN, ABS, SIN & COS. The users can create as many functions as they want.

For the connection definition we have:

"connection_genes":
		[
			{"exist": 1,"innovation": 0,"in": 0,"out": 2,"weight": 1.,"enable": 1},
			{"exist": 1,"innovation": 1,"in": 1,"out": 2,"weight": 1.,"enable": 1}
		]