Epoch vs Iteration when training neural networks

0 votes

What is the difference between epoch and iteration when training a multi-layer perceptron?

Feb 10 in Machine Learning by surbhi
• 3,810 points
100 views

1 answer to this question.

0 votes
In the language of neural networks, this is what it means:

One epoch equals one forward and one backward pass through all of the training scenarios.
Batch size refers to the amount of training examples that can be processed in a single forward/backward pass. You'll need additional memory space as the batch size grows.
The number of iterations equals the number of passes, with each pass containing [batch size] instances. To be clear, a pass consists of one forward and one backward pass (we do not count the forward pass and backward pass as two different passes).
For example, if your batch size is 500 and you have 1000 training instances, 1 epoch will take 2 iterations to complete.

Some people use the term "batch" to refer to the full training set, while others use it to refer to the number of training instances in one forward/backward pass (as I did in this answer). The term mini-batch can be used to avoid misunderstanding and clarify that batch refers to the amount of training instances in a single forward/backward pass.
answered Feb 10 by CoolCoder
• 4,360 points

Related Questions In Machine Learning

0 votes
1 answer
+1 vote
0 answers

text mining new set of data in production environment expect training feature

Hi, I have trained a model based on ...READ MORE

Nov 28, 2019 in Machine Learning by MANOJ
• 130 points
247 views
0 votes
1 answer

tensorflow is installed but when importing in pycharm gets an error

Hi@soniya, This problem is related to the version. ...READ MORE

answered Sep 14, 2020 in Machine Learning by MD
• 95,360 points
2,298 views
0 votes
2 answers

How do neural networks used in AI and deep learning learn?

A neural network is a hardware or software ...READ MORE

answered Apr 8, 2019 in Others by SA
• 1,050 points
378 views
0 votes
1 answer

Role of the bias in neural networks.

Hi@akhtar, The activation function in Neural Networks takes ...READ MORE

answered Jul 15, 2020 in Machine Learning by MD
• 95,360 points
704 views
0 votes
1 answer
webinar REGISTER FOR FREE WEBINAR X
Send OTP
REGISTER NOW
webinar_success Thank you for registering Join Edureka Meetup community for 100+ Free Webinars each month JOIN MEETUP GROUP