Home

Welcome to the GeneticMNIST wiki!

Here I can elaborate a bit about the Technique used to create this simulation.

The result, a pure C++ fast execution code is generated out of the language Cortex Assembler which is a opcode intermediate language for genetic codes. The genetic codes do have two types of parameters. One type is the layout of opcodes that defines a network topology. By ordering different opcodes after each other different routing of information can be achieved in a network graph. Just like pipelines of information in various directions. These instruction opcodes is called the network topology DNA (inspired from A,C,G,T) building bricks in biological DNA

The second type is the parameter DNA. They are parameters used by the network dna to control execution flow and values. The parameter DNA can shut down a pipeline totally and stop the flow of information. The parameters can multiply, divide, etc..

The two sets of parameters can be seen as one unit. By imagine a infinite network with nodes all interconnected, and then applying an infinite number of parameters where almost everyone is set to zero and then take away all redundant information and only keep the non zero parameters and the network DNA that defines the routing for the non zero parameters, you end up with a condense description of a network that actually is infinite.