Fill NTuple for MLP

ROOT
1

Hi,
I am trying to create a Neural Network with 576 st input variables and one output to identify a signal in a picture. From the tutorials “mlpHiggs.C” and “MLPRegression.C” I found that you input the training data by command

train = ROOT.TNtuple("data","data","X:type")

But here I am a bit unsure how to declare that “X” is suppose to be a vector with 576 st elements. So far I have tried to store my training data with the line.

train.Fill(x,y)

But I am unable to have “x” in the correct format. It dose not accept numpy arrays, list or strings. Dose some one know what format I must have on the “x vector” in order to use “.Fill”.

Thanks,
-Fredrik

2

Hi Fredrik,

the TNtuple is designed to store rows of individual single precision floating point numbers.
I see two solutions. The former is to switch to a TTree, which will allow you to store stl vectors in its branches and the latter is to have 576 columns in the TNtuple itself.

Cheers,
Danilo

3

Thanks for response Danilo,

And your advice sounds good, I have not tried the solution you mention since other solution came which drastically reduced my number of input and had promising results. I might get back to this specific example later though, I post in that case.

-Fredrik