Final merge before the first release

Final merge before the first release

README.rst

@@ -32,7 +32,9 @@ MiraPy can be used for problem solving using ML techniques and will continue to
 - Astronomical Image Reconstruction using Autoencoder
 - Classification of the first catalog of variable stars by ATLAS
 - HTRU1 Pulsar Dataset Image Classification using Convolutional Neural Network
-- Curve Fitting using Autograd (incomplete implementation)
+- Variable star Classification using Recurrent Neural Network (RNN)
+- 2D visualization of feature sets using Principal Component Analysis (PCA)
+- Curve Fitting using Autograd (basic implementation)
 
 There are more projects that we will add soon and some of them are as following:
 

mirapy/autoencoder/models.py

@@ -1,10 +1,14 @@
-from keras.models import Model
+import os
+from keras.models import Model, load_model
 from keras.layers import *
 import matplotlib.pyplot as plt
 
 
 class Autoencoder:
     def __init__(self):
+        """
+        Base Class for autoencoder models.
+        """
         self.model = None
         self.history = None
         self.dim = None
@@ -13,19 +17,41 @@ def __init__(self):
         self.decoded = None
 
     def compile(self, optimizer, loss):
+        """
+        Compile model with given configuration.
+
+        :param optimizer: Instance of optimizer.
+        :param loss: String (name of loss function) or custom function.
+        """
         pass
 
     def train(self, x, y, batch_size=32, epochs=100, validation_data=None,
               shuffle=True, verbose=1):
+        """
+        Trains the model on the training data with given settings.
+
+        :param x:  Numpy array of training data.
+        :param y:  Numpy array of target data.
+        :param epochs: Integer. Number of epochs during training.
+        :param batch_size: Number of samples per gradient update.
+        :param validation_data: Numpy array of validation data.
+        :param shuffle: Boolean. Shuffles the data before training.
+        :param verbose: Value is 0, 1, or 2.
+        """
         pass
 
     def predict(self, x):
-        pass
+        """
+        Predicts the output of the model for the given data as input.
 
-    def evaluate(self):
+        :param x: Input data as Numpy arrays.
+        """
         pass
 
     def plot_history(self):
+        """
+        Plots loss vs epoch graph.
+        """
         plt.plot(self.history.history['loss'])
         if 'val_loss' in self.history.history.keys():
             plt.plot(self.history.history['val_loss'])
@@ -35,18 +61,42 @@ def plot_history(self):
         plt.legend(['Train', 'Test'], loc='upper right')
         plt.show()
 
-    def load_model(self, path, model_name):
-        pass
-
-    def save_model(self, path, model_name):
-        pass
+    def save_model(self, model_name, path='models/'):
+        """
+        Saves a model into a H5py file.
+
+        :param model_name: File name.
+        :param path: Pa
+        """
+        path += model_name
+        self.model.save(path)
+
+    def load_model(self, model_name, path='models/'):
+        """
+        Loads a model from a H5py file.
+
+        :param model_name: File name.
+        :param path: Pa
+        """
+        path += model_name
+        if os.path.exists(path):
+            self.model = load_model(path)
+        else:
+            raise FileNotFoundError("Model does not exists")
 
     def summary(self):
         pass
 
 
 class DeNoisingAutoencoder(Autoencoder):
     def __init__(self, img_dim, activation='relu', padding='same'):
+        """
+        De-noising Autoencoder used for the astronomical image reconstruction.
+
+        :param img_dim: Set. Dimension of input and output image.
+        :param activation: String (activation function name).
+        :param padding: String (type of padding in convolution layers).
+        """
         self.dim = img_dim
         self.input_img = Input(shape=(*img_dim, 1))
 
@@ -70,11 +120,28 @@ def __init__(self, img_dim, activation='relu', padding='same'):
                               padding=padding)(x)
 
     def compile(self, optimizer, loss):
+        """
+        Compile model with given configuration.
+
+        :param optimizer: Instance of optimizer.
+        :param loss: String (name of loss function) or custom function.
+        """
         self.model = Model(self.input_img, self.decoded)
         self.model.compile(optimizer=optimizer, loss=loss)
 
     def train(self, x, y, batch_size=32, epochs=100, validation_data=None,
               shuffle=True, verbose=1):
+        """
+        Trains the model on the training data with given settings.
+
+        :param x:  Numpy array of training data.
+        :param y:  Numpy array of target data.
+        :param epochs: Integer. Number of epochs during training.
+        :param batch_size: Number of samples per gradient update.
+        :param validation_data: Numpy array of validation data.
+        :param shuffle: Boolean. Shuffles the data before training.
+        :param verbose: Value is 0, 1, or 2.
+        """
         self.history = self.model.fit(x, y,
                                       epochs=epochs,
                                       batch_size=batch_size,
@@ -83,9 +150,21 @@ def train(self, x, y, batch_size=32, epochs=100, validation_data=None,
                                       verbose=verbose)
 
     def predict(self, x):
+        """
+        Predicts the output of the model for the given data as input.
+
+        :param x: Input data as Numpy arrays.
+        """
         return self.model.predict(x)
 
     def show_image_pairs(self, original_images, decoded_images, max_images):
+        """
+        Displays images in pair of images in grid form using Matplotlib.
+
+        :param original_images: Array of original images.
+        :param decoded_images: Array of decoded images.
+        :param max_images: Integer. Set number of images in a row.
+        """
         n = min(max_images, len(decoded_images))
 
         plt.figure(figsize=(20, 8))