Output details

Unification neural networks : unification by error-correction learning

<11> The paper shows that error-correction learning observed in neural networks has direct correspondence to the famous algorithm of first-order unification (by Robinson). The paper uses this correspondence to suggest a neural-network interpreter for first-order unification. It is produced as an output of Komendantskaya’s EPSRC Postdoctoral Research fellowship EP/F044046/1, EP/F044046/2, (total award £366.115, years 2008 – 2011). Prior to this journal publication, the preliminary results were published in conferences NeSy’08, LATA’07, ICNC’09, STP’09 and presented as invited talks in the following universities: St Andrews, Bath, Imperial College, Stirling, Aberdeen, INRIA Sophia-Antipolis (France), Osnabrueck (Germany).