Associative Memory
- I'm hoping to start working on Associative Memory shortly.
- Here I'm talking about a semantic net like memory.
- We can encode a lot of words based on existing semantic hierarchies
like word net.
- This will give us persistent hierarchical stable states (CAs).
- We can select a series of relationships and use existing text
to learn these relationships.
- For example we could try to learn the cause relationship.
- We could find text like "The boy broke the window." to learn that
boys cause windows to break.
- There will be a lot of complexity.
- We'll test it as a priming cognitive model.
- This won't give full fledged semantics; it's not learning
from the environment. It will give us a nice
entry for semantics for a lot of words.
- This will take advantage of large SpiNNaker networks.
- The resulting system should be readily plugged into a conversational
agent.
- This should provide a nice subsystem that will be used by the
architecture agent. This can provide semantics for a number of simple
relationships like "support".