Summary form only given, as follows. A neural blackboard architecture is proposed in which a high-order associative memory network serves as a global shared memory for one or more feedforward-type neural inference engines to allow the system to reason in a time-delayed, or sequential, manner. New learning algorithms are proposed to speed up slow learning and to partially alleviate the tough credit apportionment problem inherent in any multistep decision.
|Original language||English (US)|
|Title of host publication||Unknown Host Publication Title|
|Number of pages||1|
|State||Published - 1987|
All Science Journal Classification (ASJC) codes