What does hidden_layer = layers.Dense(100, activation="relu")(input_layer) do?

Question

I saw the following in a guide to text classification:

hidden_layer = layers.Dense(100, activation="relu")(input_layer)

What does the (input_layer) mean?

The context is:

def create_model_architecture(input_size):

    # create input layer 
    input_layer = layers.Input((input_size, ), sparse=True)
    
    # create hidden layer
    hidden_layer = layers.Dense(100, activation="relu")(input_layer)
    
    # create output layer
    output_layer = layers.Dense(1, activation="sigmoid")(hidden_layer)

    classifier = models.Model(inputs = input_layer, outputs = output_layer)
    classifier.compile(optimizer=optimizers.Adam(), loss='binary_crossentropy')
    return classifier

Also - is this a sequential model?

you need to define the input layer, so when you are compiling the model with `models.Model()` you would be able to specify it. no, it used functional API. — SoheilStar, May 08 '21 at 03:21
Please use a wonderful piece of technology called "search engine": https://keras.io/guides/functional_api/ — desertnaut, May 08 '21 at 09:19

score 0 · Answer 1 · answered May 08 '21 at 03:22

0

See input layer is nothing but how many neurons or nodes you want for input. Suppose I have 3 features in my dataset then I'll have 3 neurons in input layer. And yes it's sequential model.

answered May 08 '21 at 03:22

Vibhav Surve

73
1
6

I think I'm confused by the syntax - why are there two sets of parentheses - ()() - after layers.Dense? – Pearl May 08 '21 at 03:58
can you please specify line number – Vibhav Surve May 08 '21 at 04:02
can you please specify line number. Edit:- As you can see line before layers.Dense() in second parentheses we specify the layer before it. – Vibhav Surve May 08 '21 at 04:09
Why is this syntax usage not mentioned in [this guide on Keras](https://keras.io/guides/sequential_model/)? How would I use `model.add()` to create a model that does the same thing? – Pearl May 08 '21 at 04:17
Yes you can . . – Vibhav Surve May 08 '21 at 04:22
In your case you are using a function to do so.Always there are quite someways you can do the same work – Vibhav Surve May 08 '21 at 04:24
@Pearl Because you cannot do the same with a Sequential model, the Functional API is meant to create graph like models, not sequential ones. – Dr. Snoopy May 08 '21 at 11:07

What does hidden_layer = layers.Dense(100, activation="relu")(input_layer) do?

1 Answers1