MNIST-Using-Jupyter-Notebook

Classification of MNIST dataset using Jupyter Notebook. There are two two types of neural network used,

• A model with One with Input layer and Ouptput Layer and,

# Creates a simple model with imput and output layer with no hidden units
model = keras.Sequential([keras.layers.Flatten(input_shape=(28, 28)),
                        keras.layers.Dense(10, activation='sigmoid')
                        ])
# Compiles the model with different parameters
model.compile(optimizer='adam',
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy']
             )
# Trains the model with features and associated labels
model.fit(X_train, y_train, epochs=5)

• Other with one additional layer i.e. Hidden Layer

# Adds one more layer to the model i.e. a hidden layer
model_hidden = keras.Sequential([keras.layers.Flatten(input_shape=(28, 28)),
                        keras.layers.Dense(50, activation='sigmoid'),
                        keras.layers.Dense(10, activation='sigmoid')
                        ])
model_hidden.compile(optimizer='adam',
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy']
             )
model_hidden.fit(X_train, y_train, epochs=5)

The code check the performance of the model and is an attempt to check the influence of hidden layer on datasets.

Experiments

The expermiment performed is based on the confusion matrix as follows,

Without Hidden layer	With Hidden layer

The performance of Deep Learning Network is better with hidden units.