Figure 4-6 Results: (a) Flow: desired in blue, predicted in red; (b) MSE criterion, full line
is for expert 1, dotted line is for expert 2; (c) Winner; (d) Pressure.
OLS clustering. Figure 4-7- shows the simulation results for a RBFN, this time
using OLS clustering for the RBF layer. The values of the error criterion are in the same
range as the ones for the k-means clustered experts, but the segmentation performed a lot
better: all mandatory breaths are detected. This is due to the fact that the criterion value
for expert 2 increases more when there is a change in regime than for the other clustering
method, therefore compensating faster for the "buildup" (the decrease of the other
criterion value has essentially the same exponential rate for all models, determined by the
memory depth).
With such encouraging results on the segmentation without adaptation, we now
consider the results for other segmentation approaches with online adaptation of the
experts.
i I I I I -
00 1000 100 2000 200 3000
500 1000 1600 2000 2500 3000