Assignment 4 - In Class Data Science Competition

The in-class data science competition will allow you to compete with other participants in the course. Please register here:

Fill this form once you have participated in the competition for us to map your past submissions and usernames from Kaggle:


This kaggle competition looks very cool, however I think atleast 30 days should be given.


I don’t understand cause the correction notebook of kaggle competition is here :

I am really excited for this competition with the time frames given. Will we be given any indication what the prizes will be?Perhaps that will incentivize and motivate great quality work lol :slight_smile:


It’s not the correction but just a start.

Could you explain what are going to on this competition. are we going to start a new notebook from the scratch or we are going to edit existing notebook? if yes where are we to modify?

1 Like

You can use the starter notebook and modify it however you need to improve the scores. The starter kernel is here


Does anybody know what is the exact deadline for submission this assignment?

Hi all,
I have shared some learning from the competition in Kaggle

Hope this helps as you build your models

for now the deadline is 29 June. Aakash did mention that it may be extended
However, best to submit on time/early rather than the last minute rush!


Tried running the sample notebook, got the following error-

1 Like

This is due to internet turned off.

I highlight this and other issues in my post; these should help and also experiences I moved to rank #3 in the competition:



I need some help. I started working with the starter notebook, but after training my val_score is 0.0000. This is weird, but could be possible if my model is really bad. But when testing on some sample images (with predict_single()) there is neither a prediction nor a label, only the picture gets shown.

Has somebody faced a similar problem or knows how to solve it?

1 Like

you should share your model, these images give us no way to know what you are doing.

" Maximum interactive GPU session count of 1 reached." How to renable the GPU. my training is too slow. Thanks

You must have a kernel running somewhere else, first close that or disable GPU in that notebook.


This is my model, it’s mostly the starter model with minor changes:

But shouldn’t the labels be shown no matter what the models look like?

I think the error is from nn.Dropout() , i think it is necessary to have some probability of dropout as a parameter.

I have written a starter’s guide to Kaggle. Please have a look if this is your first experience with Kaggle.


I tried it without the dropout but the problem is still there. I also took a look in the documentation, if no number is specified, the default value of Dropout is 50%.
Nevertheless thank you for your help! :slightly_smiling_face: