tf.keras.Input(shape=(400,)), #specify input shape

Course Q&A Machine Learning Specialization Advanced Learning Algorithms
1

Here S input = (2000, 400)

However, we initialise Keras inputs as,

tf.keras.Input(shape=(400,)), #specify input shape

Shouldn’t 400 be specified as column as opposed to row vector here?

2

Hello @Purnima_Sethi,

Tensorflow needs us to specify the shape of a sample, so it is just (400, ).

Raymond

3

400 is the number of features.

When we define the layers in Keras, we don’t specify the number of examples, only the number of features. That’s why it’s (400,)

4

Thank you. I assumed it would be (: , 400) since the training examples are unspecified.