:D
Not unless I accidentally make skynet! (my application is for theoretical interest of merging AI/neural networks and control algorithms to get the "best of both worlds", ie, a state space that doesn't require mathematics, and a convolutional network that is physically viewable in the n-dimensional state space.
I came up with a new idea of "strings" in the n-dimensional matrix state space, ie a sequence of states that define a set or subset of state space states. These can be gathered from training (reading sensor and I/O) and filling in states to build a string. They can also be labelled as a certain function, ie a string that does a rotation while holding other variables constant.
The strings could be mathematically combined to do higher level functions, ie rotate while translating.
If the state space is filled in with training, then given the current state space, the closest match to a filled in state space can be found and then the string that state belongs to can be traversed or another string can be translated to in the n-dimensional space.
Pretty abstract, but I am trying to come up with a concept that can function as a convolutional neural network, and still be describable and understandable physically, like a mathematical state space system is.
cheers, Jamie