I want to ask about the solutions. I have read the overview of challenge that " Your job will be to build a machine learning or deep learning model that can intake a seismic line jigsaw puzzle and correctly put the piece together in the correct order".
I wonder whether I can intergrate AI model with another search algo, genertic algo or use Reiforcement Learning?
Hello @vutuantrinh2000
Yes, we agree that integrating algorithms on the edge of machine learning and artificial intelligence could lead to an interesting and impactful solution. Feel free to develop a combination of techniques to enhance the performance of your model.
Best of luck with your approach,
Onward Team
Hello,
Just curious to know if it is accepted to use only generative AI such as ChatGPT ?
Hello @motoki.saitama
We’re not entirely certain what specific approach you’re referring to when mentioning ‘just generative AI like ChatGPT’. However, using probabilistic and autoregressive techniques from generative AI can help improve the predictive and decision-making capabilities of a puzzle-solving system.
Kind regards,
Onward Team
Hello,
I mean we could try to downsample the patches and let chatgpt figures out how to put them together. Well, it is more prompt engineering rather than machine learning.
BTW, is it possible that you separate each image in a json file? the train.json is too big to be parsed by a 64 GB memory computer.
It sounds like you have an interesting approach for solving the challenge. Give it a try and see how the resulting submission is scored on the Predictive Leaderboard. As for the memory challenges, try using the memory map (mmap) library to load data, the implementation can be find in our starter notebook.
Good luck!
Onward Team