AutoEncoder

Histogram classfier based on the MSE of an autoencoder reconstruction

The AutoEncoder derives from the generic HistogramClassifier.
For this specific classifier, the output score of a histogram is the mean-square-error (MSE) between the original histogram and its autoencoder reconstruction.
In essence, it is just a wrapper for a tensorflow model.

---

---

[class] AutoEncoder

comments:

histogram classfier based on the MSE of an autoencoder reconstruction  
the AutoEncoder derives from the generic HistogramClassifier.   
for this specific classifier, the output score of a histogram is the mean-square-error (MSE)   
between the original histogram and its autoencoder reconstruction.  
in essence, it is just a wrapper for a tensorflow model.  

⤷ __init__

full signature:

def __init__( self, model=None, modelpath=None )  

comments:

intializer from a tensorflow model  
input arguments:  
- model: a valid tensorflow model;  
         it does not have to be trained already,  
         the AutoEncoder.train function will take care of this.  
- modelpath: path to a stored tensorflow model,  
             it does not have to be trained already,  
             the AutoEncoder.train function will take care of this.  
note: model and modelpath are alternative options, they should not both be used simultaneously.  

⤷ train

full signature:

def train( self, histograms, doplot=True, epochs=10, batch_size=500, shuffle=False, verbose=1, validation_split=0.1, **kwargs )  

comments:

train the model on a given set of input histograms  
input arguments:  
- histograms: set of training histograms, a numpy array of shape (nhistograms,nbins)  
- doplot: boolean whether to make a plot of the loss value  
- others: see the keras fit function  
- kwargs: additional arguments passed down to keras fit function  

⤷ evaluate

full signature:

def evaluate( self, histograms )  

comments:

classification of a collection of histograms based on their autoencoder reconstruction  

⤷ reconstruct

full signature:

def reconstruct( self, histograms )  

comments:

return the autoencoder reconstruction of a set of histograms  

⤷ save

full signature:

def save( self, path )  

comments:

save the underlying tensorflow model to a tensorflow SavedModel or H5 format.  
note: depending on the extension specified in path, the SavedModel or H5 format is chosen,  
      see https://www.tensorflow.org/guide/keras/save_and_serialize  

⤷ load

full signature:

def load( self, path, **kwargs )  

comments:

get an AutoEncoder instance from a saved tensorflow SavedModel or H5 file  

---