Table 3.

Details of the training procedure used for each of the benchmarking algorithms tested. In all cases the parameters values listed were those found to produce the best results. Parameter values were kept constant across variations in the task. All algorithms except $\text{nmfsc}$ use an online learning procedure and hence training time is measured in cycles, whereas for $\text{nmfsc}$ training time is measured in epochs. See the caption of Table 2 for further details.

Algorithm	Training time	Iterations	Weight initialisation	Parameter values
$\text{fyfe2}$	200 000 cycles	n/a	mean = $\frac{1}{4}$, std = $\frac{1}{32}$	β = 0.01, ν = 0.1
$\mathrm{harpur}2$	20 000 cycles	50	mean = $\frac{1}{32}$, std = $\frac{1}{8}$	β = 0.0025, μ = 0.05, θ = 0.5
$\mathrm{nmfsc}$	2 000 epochs	n/a	mean = $\frac{1}{2}$, std = $\frac{1}{8}$	s W = 0.5, s Y = none
$\text{di}$	20 000 cycles	25	mean = $\frac{1}{36}$, std = 0.001	β + = 0.25, β − = 0.25
$\text{dim}$	20 000 cycles	50	mean = $\frac{1}{16}$, std = $\frac{1}{64}$	β = 0.05