CODEWITHSUNDEEP
pythontensorflowmachine-learningdeep-learningtflearn
neel
neel

Reputation: 752

What does the second argument of tflearn.fully_connected represent?

I am trying to learn tflearn. But I have a few doubts.

In the following line

net = tflearn.input_data(shape=[None, len(train_x[0])])

is len(train_x[0]) the shape of my output matrix? If not, what is it?

Second doubt is: what is 8 in this line?

net = tflearn.fully_connected(net, 8)

I tried to search and I found it is n_units, but what are they, and how am I supposed to choose how many units I will require in which case?

Upvotes: 2

Views: 2087

Answers (1)

Nicki Skafte
Nicki Skafte

Reputation: 489

The line

net = tflearn.input_data(shape=[None, len(train_x[0])])

means that tflearn expects that the input to your network has shape [?, len(train_x[0])]. In your case, I think that train_x is a matrix, meaning that len(train_x[0]) would give you the number of columns in your matrix.

If you look at the documentation for tflearns fully connected layer (http://tflearn.org/layers/core/), you will see that the 8 corresponds to the n_units argument

net = tflearn.fully_connected(net, 8)

meaning this line will create a fully connected layer with 8 hidden units/neurons.

Upvotes: 1

Related Questions