Module Source.data

import numpy as np
import matplotlib.pyplot as plt
import pickle


def batch_normlize(x):
   """ This function is used to normalize an array/vector of data

   Parameters:
     
      x (array/vector): Array/vector of unnormalized numbers

   Returns:
     
      (array/vector) : normalized array/vector where normalized x = (x - mean) / np.sqrt(variance + eps)

   """
   eps=0.00001
    
   mean = x.mean(axis=0)
   variance = x.var(axis=0)
    
   xnormalized = (x - mean) / np.sqrt(variance + eps)
 
   return xnormalized


def flatten(x):
   """ This function is used to reshape a matrix into a flat shape.
     usefull in preproccessing of input data

   Parameters:
     
      x (matrix/vector): Multi-dimension matrix/vector

   Returns:
    
      (matrix/vector): flatted matrix where each example is in a single row or column

   """
   return x.reshape(x.shape[0], -1).T

def onehot(y_train):
   """ This function is used to change data in the normal form into one-hot form 

   Parameters:
    
      y_train (matrix): Matrix of numbers

   Returns:
     
      (matrix) : One-hot form of the data

   """
   y = np.zeros((y_train.size, y_train.max()+1))
   y  [np.arange(y_train.size),y_train] = 1
   y = y.T
   return y



class visualize:

   points_y=[]
   points_x=[]
     
   fig=plt.figure(figsize=[10,8])
   graph=fig.subplots()

def __init__(self,title,y_label,x_label):
   """ visualization class to draw live cost after training, to initialize an instance from this class supply 3 parameters

   Parameters:
    
      title (string): title of the graph
      y_label (string): label of y coordinate
      x_label (string: label of x coordinate

   """
     
   self.graph.set_ylabel(y_label)
   self.graph.set_xlabel(x_label)
   self.graph.set_title(title)
            
def addpoint_y(self,pointy):
   """ This function is used to add a new point to be drawn on a graph 

   Parameters:
     
      pointy (float): value of y coordinate 

   Returns:
     
      graph: graph drawn after adding new point

   """
   self.points_y.append(pointy)
          
   if (len(self.points_x) ==0):
      self.points_x=[1]
               
   else:
      self.points_x.append(len(self.points_x)+1)
               
   self.graph.plot(self.points_x,self.points_y,'bo')
   self.fig.show()
   plt.pause(0.0001)

            
def save(filename,model):
   """ This function saves the model ( all of it's parameters, weights, losses ,... ) in an external file 

   Parameters:
      
      filename (string): Path of file to be saved in, filename must be ".sav" type.
      model    (object): "model" required to be saved.

   Returns:
      (None)

   """
   pickle.dump(model, open(filename, 'wb'))

def retrive(filename):
   """ This function loads a saved model ( all of it's parameters, weights, losses, ...) from external file

   Parameters:
      
      filename (string): Path of file to be retrived, filename must be ".sav" type.

   Returns:
      
      (object): Object of model that saved before as whole with its structure(Layers,Nodes,Activation functions)

   """
   return pickle.load(open(filename, 'rb'))