We’re having a heat wave here in the UK so as I lay out under my lawn sprinkler (suitably primed with a mixture of spring water and a rather nice chardonnay – me that is, not the lawn sprinkler) I began leafing through one of my favorite journals  - Biological Cybernetics. Yes, I know you all have a prized copy hidden away somewhere – under the bed or in the garden shed – but there, back in the 1981 volume, I came across Associative Search Network: A Reinforcement Learning Associative Memory introduced by the authors as:

In this paper we describe an associative memory structure which is not told by some outside process (e.g., a “teacher”) what pattern it is to associate with each key. Instead, for each key, the network must search for that pattern which maximizes an external payoff or reinforcement signal. As this kind of learning proceeds, each key causes the retrieval of better choices for the pattern to be associated with that key.

and I wondered what relevance this might have to XBRL. Read the rest of this entry »