Figure 4
From: Stable Odor Recognition by a neuro-adaptive Electronic Nose
Rates of correct classification for two fault conditions.
A) Correct classifications with adaptive network (red) or without adaptive network (gray) depending on which sensor exactly was in fault. Red and gray horizontal lines are the control cases (i.e., without fault) and show the average correct classification rates with adaptive network (red) or without adaptive network (gray). B) Same as in A but for one randomly fluctuating sensor output. Note that in cases where the faulty sensor contributed more noise than the other sensors, the classification rate of the network-PLS-DA can exceed the maximum.
