# TMVA TMultiLayerPerceptron => plateau searches

Hi all,

Has someone already tried to use one of the TMVA classes or the TMultiLayerPerceptron to find a plateau…

Input (x , y , error_y ) \in { ( 1 , 10 , 0.1 ) ( 2 , 9 , 0.2 ) ( 3 , 8 , 0.25 ) ( 4 , 8.1 , 0.30 ) ( 5, 7.8 , 0.29 ) ( 6 , 7.7 , 0.30) }

Output => the plateau is { (x,y) , x \in [[3,6]] }

Of course, in my experience, absolute value of derivative < epsilon is not good enough… because signal can be noisy (error_y not so small) …