/NeuralNetworkForHistograms

Primary LanguagePython

NN for histograms

nn

Input features size = 111
Hidden Layers = 1 X 55

Classifying histograms with a Neural Network.

Histograms are represented as numpy arrays where each elem is number of Points in each [Yi, Yi + dy] (or each bin). We have 111 Features and one output (Binary Classification).

Why Kears ?

  • Sklearn MLPClassifier module is poor compared to APIs like Theano.
  • Keras is well documented and very user friendly. In fact, if you are already familiar with sklearn classifiers, you wont be disorientated.
  • Keras allows you to save models and load them. Models are stored as JSON and YAML files which is perfect.
  • Keras exemples in officiel GitHub are amazing and very helpful.