What does "loss" in the output from training indicate?

Course Q&A TensorFlow: Advanced Techniques Specialization Custom Models, Layers and Loss Functions with TF
1

Hi!
I have enrolled in the first course and I saw the code of training a model on energy efficiency data and noticed that there are three losses in the output named: loss, y1_output_loss, y2_output_loss.

what does loss indicate?
Thanks

2

Hi @ibrahim97,

Could you share where you saw the loss? The paper/ implementation would give a better idea of what the different losses were doing!

3

@thearkamitra thank you for your reply.

I saw the loss in the Multi-output code walkthrough video at 4:30

this is an screenshot for the loss from the video

image
image1920×1080 95 KB

You can see that there are three names for the losses: loss, y1_output_loss, and y2_output_loss

4

Hi @ibrahim97,

Yes I understand now. If you go to the corresponding lab on Week 1 of Course 1, you will find that there are two outputs of the model. Along with that, the loss if the square loss. The y1_loss is the loss from y1 output and y2_loss is that from the y2 output. The total loss is the sum of the two losses.

There are five outputs in total, 3 of which were explained in the paragraph above. For the other two losses, they are the mse loss from y1 and y2 respectively.

Hope this helps!

2 Likes
5

okay thank you so much