Mathematical justification of recurrent neural networks with long- and short-term memories
J.T. Lo, Devasis Bassu · 2003
Two theorems are given, that justify the use of multi-layer perceptrons with interconnected neurons (MLPWIN) with long- and short-term memories (LASTMs) for both L/sub p/ and risk-sensitive adaptive processing. The benefits of using a MLPWIN with LASTMs include less online computation, no poor local extrema to fall into, and much more timely and better adaptation.