Atra - Associative Trace Architecture

  From Hebb, Hopfield, and Associatron to Atra Rediscovering Associative Memory for First-Person Autonomy Atra did not appear suddenly. For me, Atra has gradually emerged through a long path: Hebb, Hopfield networks, Associatron, non-monotonic associative memory, and finally the need for first-person autonomy. Atra is not simply an AI agent. It is not a system where an LLM gives commands to a robot. It is not a model designed only to improve classification accuracy by using correct labels. What I am trying to study with Atra is a way to extend associative memory toward first-person autonomy. To explain why, I first need to look back at the flow from Hebb to Hopfield and Associatron. Then I need to explain why non-monotonicity, difference, carry, and dream-like slack become necessary. Hebb's idea One starting point is Hebb's idea. In simple terms, it can be described as follows: Things that are active together become connected. A familiar form of this idea can be written as: Del...

  Atra is my independent research project at C-side Laboratory, Japan. The name Atra means Associative Trace Architecture. This research started from Dr. Kaoru Nakano's Associatron . I have been interested in the Associatron for a long time because it is not just a memory model. It has a very simple but important idea. A part can call the whole. A small cue can recall a larger memory. A pattern can connect to another pattern. A memory is not stored in one fixed place, but distributed across the whole system. This is very different from ordinary computers. A conventional computer usually needs an address. It needs to know where the data is. Associative memory begins in a different way. It begins from a cue. That difference is very important to me. Why Associatron still matters Dr. Nakano's Associatron was proposed in 1972. It stores entities as distributed patterns and recalls the whole from a part. If the cue is large, the recall becomes more accurate. If the cue is small, the ...