Basic autoencoders

Let's look at a basic example of an autoencoder that also happens to be a basic autoencoder. First, we will create an AutoEncoder class and initialize it with the following parameters passed to __init__():

• num_input: Number of input samples
• num_hidden: Number of neurons in the hidden layer
• transfer_function=tf.nn.softplus: Transfer function
• optimizer = tf.train.AdamOptimizer(): Optimizer

Note

You can either pass a custom transfer_function and optimizer or use the default one specified. In our example, we are using softplus as the default transfer_function (also called activation function): f(x)=ln(1+ex).

Autoencoder initialization

First, we initialize the class variables and weights:

 self.num_input = num_input
 self.num_hidden = num_hidden
 self.transfer = transfer_function
 network_weights = self._initialize_weights()
 self.weights = network_weights

Here, the _initialize_weigths() function is a local function that initializes values for the weights dictionary:

• w1 is a 2D tensor...