deeplearning-numpy

Simple deeplearning library using numpy only.

Usage

call nn.model with x and x with keyword parameters

parameters

• X : input to the model - (no. of features x no. of samples)
• Y : output of the model - (no. of classes x no. of samples)

Keyword parameters

• alpha : Learning Rate
  default : 0.01
• iter : Iterations
  default : 3000
• hidden_layer_dims : Hidden layer dimentions, also decides the number of hidden layers based on the length of this list default : []
• activation : Activation function for the other layers and the last layer as a list of length 2.
  supports : sigmoid, tanh, relu, leaky_relu, softmax
  default : ['tanh','sigmoid']
• batch_size : Mini batch size
  default : X.shape[1]
• dev_set_ratio : Dev set to total data-set ratio.
  default : 0.02
• parameters_file : File-name for the parameters file to import incase of using/training a pretrained model.
  default : None