Table 1.

NL vs. ANN — a comparison of properties

Properties	ANN	NL	Remarks
Neuron	Linear followed by a	Non-linear and chaotic	Chaos allows for a rich set
	nonlinear activation		of properties to be exploited.
Output of a	Scalar	Variable length vector	Neurons in NL perform non-linear
neuron			computations as compared with simple
			weighted linear addition in ANN.
Universal	Satisfies UAT	Satisfies UAT	NL satisfies UAT with an exact
approximation			specification on the number of neurons
theorem			needed for approximating a real-valued
(UAT)			discrete-time function with finite support.
Activation	Yes	No	The nonlinearity in ANN is provided by the
functions			activation function which is not needed for NL.
Backpropogation	Yes	No	Not currently used. NL could employ
			backpropagation in the future if needed.