Deep Learning Interview Series #7-Asked In Interview-Epochs Vs Batch Vs Iterations In Deep Learning

We at iNeuron are happy to announce multiple series of courses. Finally we are covering Big Data,
Cloud,AWS,AIops and MLops. Check out the syllabus below.


Incase of any queries you can contact the below number. Happy Learning!!

8788503778
6260726925
9538303385
8660034247
9880055539

[hardikvegad]

Epoch: One iteration over the entire dataset.

Batch: When we cannot pass the entire dataset we split them into batches.

Iteration: If we have 1000 images as data and a batch size of 20, then an epoch will run 1000/20 = 50 iterations.

[SanjeevKumar-xlro]

Thank you so much! Your short videos are incredibly helpful. I’m an IT professional with 20 years of experience, currently working on transforming my career into the field of Data Science

[mukeshkumaryadav]

A very good explanation of epoch, batch and iteration. Thank you so much
epoch: No of time we want our data to repeatedly pass to model training.
batch: We don't actually pass all data at once rather we pass in a batch size. So entire data is internally divided in groups that is called batch and then we pass each batch one by one. Once all batches are passed we say one epoch is finished.
Iteration: passing data in batches in model is called iteration.

[AAND]

V well explained, just to add a little there can be continuing linked questions like system complexity limitations, what specific algo can be used, why use it, and the maths behind it, thanks again.

[aination]

Question - In epoch 2, do we have the same records in iteration 1, or the records are shuffled and the iteration 1 in epoch 2 has different set of records than iteration 1 in epoch 1?

[KiranSharma-eyxp]

Will it update weights once per batch size ?
If yes than weights are average of batch size loss or the last data point loss ?

How many times weights update for below scenario ?
1) Whole data at once
2) Batched data

Is there any 3rd option as we give data one by one ?

[RutujaKokate-uh]

You did not mention when does the weights get updated, after every iteration?

[BlochSphere]

What would happen if the dataset size (number of images in this case) is not divisible by the chosen batch size?

[pratikbhansali]

Sir please try to ask some advanced deep learning questions

[raginibhayana]

in every epoch all the dataset will be covered? wont this lead to over fitting?

[dulangikanchana]

hi can you do a video on what is feature space?

[mansvisonawane]

Can you please help one of my friends save her father? I have messaged you on linkedin. I cannot post the link here coz youtube does not allow that. Any ways her name is Debadrita Dey. Her father has been diagnosed with lung cancer. He is the only breadwinner. Please help her reach maximum number of people. You can find her post on linkedin as well.

[muhammadusmanbutt]

Nice but I've a question? Why would we need 20 epochs of we have trained our images in epoch 1????

[pritha]

In different epochs are the batches shuffled ?

[minma]

Watch it while it is still hot from the oven.