Action Map with Regards to Deep Learning
Greetings,
I have been going over the ideas of evaluating data using artificial neural networks in many different ways, and wanted to attempt a structured approached, and not only did I want to structure the data I wanted to structure the behavior of the network. I thought of making a network of nodes connected by links, where as all links and nodes have actions; the nodes will have attributes. There will be an evaluator with a artificial neural network and attributes, along with a array of visitor pointers for all actions at each link and node that will be passed to Accept when invoking the action of the object. Each link also has conditions to access if it is enabled, or capable, of using that link for transitioning.
The evaluator will use each of the possible connected nodes attribute and the attributes of the evaluator to obtain a evaluated values that rates that linkage. I am not positive at this moment if I want to use fuzzy logic, simply the highest value or probability to determine the appropriate link to use. Any case, I will see what inputs generate what outputs during my study of this system.
You can use a symbol table with sequences that have symbols for the actions to store the behavior of the network. In this way you can replay the pattern of behavior the network produces. This would be a great feature to have and a very optimized approach to reproducing periods of time in respect to actions. This is all for the moment and thank you for your time.
Charles Alan Martin CPhT
Discussion