This feature of TensorFlow allows us to create an algorithm that accepts data and knows something about the shape of the data without knowing the amount of data going in. When we insert batches of data in training, we can easily adjust how many examples we train on in a single step without changing the entire algorithm:

# numFeatures is the number of features in our input data.
# In the iris dataset, this number is '4'.
num_explanatory_features = train_input_values.shape[1]

# numLabels is the number of classes our data points can be in.
# In the iris dataset, this number is '3'.
num_target_values = train_target_values.shape[1]



# Placeholders
# 'None' means TensorFlow shouldn't expect a fixed number in that dimension
input_values = tf.placeholder(tf.float32, [None, num_explanatory_features]) # Iris has 4 features, so X is a tensor to hold our data.
output_values = tf.placeholder(tf.float32, [None, num_target_values]) # This will be our correct answers matrix for 3 classes.