A Dynamical Systems Reformulation of the Normalized Recurrence Algorithm
Bo Pedersen · eScholarship (California Digital Library) · 2011
A Dynamical Systems Reformulation of the Normalized Recurrence Algorithm Bo Pedersen Rosetta Stone Labs Abstract: The Normalized Recurrence Algorithm is a kind of localist attractor network describing the temporal dynamics in continuous and recurrent information integration emerging in experimental psychology data (The Con- tinuity of Mind, Michael Spivey, 2007). Despite the fact that this algorithm successfully models time series data, it is somewhat unsatisfactory to deal with an algorithm within a dynamical systems context and furthermore it is difficult to prove conjectures, so we suggest this description: E(t+1)=E(t)+E(t)xC*E(t), where t is time, E is all the different activity vectors pooled together and C is a connectivity matrix (x and * is the matrix and hadamard product respectively). The normalization is performed post hoc only, and with this reformulation we can now prove the set size/convergence linearity hypothesis (p220), and reject the indirect crosstalk hypothesis (p103), and more importantly open up this model to comparisons within the field of dynamical systems where it truly belongs.