Suggestions for ML Challenge?

Suggestions for ML Challenge?
0

Hello all.

I am preparing to host an event focusing on Machine Learning, and am wracking my brain trying to think of a clever topic/topics to design a challenge that needs solving through the use of machine learning.

I am wanting a task that can be completed in under 2 hours by most coders, and it should be in the realm of an engineering problem - whether that be biomedical, mechanical, or otherwise.

I have already done an event on Deep Learning. So, I would prefer a challenge not concerning image classification.

Does anyone have suggestions for the task/s and/or datasets?

You can always take a look at this massive list of datasets here. Another place you can take a look Kaggle.

Both are great for different reasons. The first one is good defined datasets. The second one has the extra benefit of usually having a good problem to solve.

The difficulty is having a compelling problem to solve. I don’t know what kind of audience you’ll have, but if you can find a compelling story about some problem you want to solve then find some data to it, the people working on it will be more engaged.

Good luck!

1 Like

@Sky020 try creating a pool to ask people which type of content they would like to learn about.

One of my favorite events when I’m visiting home at the right time of year is an event where attendees spend a weekend giving NPO data the once over. You get to help out a good cause and get some practice with dirty, ugly real-world data. See if there’s anyone who could use a little ML help that would be willing to hand over some data to be scienced.

Thank you, eric. Those links are massively helpful. My audience will mostly be postgraduate students studying engineering disciplines. However, seeing as my event is open to all, I would hope to have a topic that is appealing enough to attract almost anyone.

I have had polls, in the past. The issue is that there is so much variance, and very little relative response (said every person who collected data). The only definitive I have is the interest in ML.

@becausealice2

I really like the idea of this. However, due to the event being sponsored by the company I work for, I cannot use their time and resources to help someone else with work. If I did manage to get permission to do this, then there would be issues with using the company name on work produced by the event attendees who are not employees.

Fair enough.

This was just posted in the data science gitter channel, might be worth a look if you still need data: https://towardsdatascience.com/google-just-published-25-million-free-datasets-d83940e24284