[frankinstien]

frankinstien, would you say that your system is "symbolic + hormones"? It would be an interesting approach.

The concept employs symbolic coding where it exploits generalized states which can represent qualities about objects, so, even chemical states such as hormones can also be generalized to various levels that relate to symbolic representations of emotions. Such emotions are associated with experiences and/or initialized concepts that when applied constrain the type of memories retrieved. So, the goal is to break down environmental information into granular components, inclusive of chemical representations that have certain effects, so even the hormones are symbolic, but they are encoded to be fuzzy.  The Object Oriented Data model has a common structure that allows it to be interrogated easily yet is very dynamic. This flexibility allows for fuzzy states to describe concepts that generalize to granular symbolism. So, for example, the symbolic state of running and the possible ranges of speed become relative to specific objects yet are a common attribute of the objects because of the symbolic term "run". This helps create implicit relationships where concepts have similarities yet each can have unique representations.

 The diagram describes an approach that capitilizes on digital infrastructures to provide a means of accessing and integrating conceptual data. By providing such an infrastructure it standardizes how information can be captured, encoded, and accessed so any kind of algorithmic process can use it, inclusive of neural networks.  Convolutional or other neural approaches, IMHO, needlessly have to configure or train a neural network to simulate such infrastructure, which they don't do very well or efficiently. This approach also allows for spontaneous learning since the symbolic encoding can also be applied to episodic events, which today is problematic for ANNs.