Z-Order Curve Clustering and Visualization
Description as a Tweet:
We created a general purpose library for doing arbitrary large dimension reduction using space filling curves. It also works the other way (1D to ND).
Inspiration:
I did dimensionality reduction techniques for data visualization over the summer and it took a really long time, I wondered why people don't use continuous space filling curves to do locality preserving visualization/clustering.
What it does:
It takes in an arbitrary amount of high dimensional data and returns a one/two/three dimensional representation on a space filling curve. It can also take one/two/three dimensional data and project it up in dimensions.
How we built it:
We spent a lot of time researching the concept and then implemented a proof of concept as well as possible in the time frame.
Technologies we used:
- Python
- AI/Machine Learning
Challenges we ran into:
Time constraints and some of my team didn't know how to program in python/use version control.
Accomplishments we're proud of:
Actually completing the project in a way that returns meaningful results.
What we've learned:
Gray Codes, Space-filing codes, Bit interleaving, Big/Little Endianess, and how to collaborate on team projects
What's next:
Making it faster (implementing it in C with bit shifting instead of slow POC string operations), allowing negative values, making it prettier.
Built with:
We made an early stage implementation in python using plotly, numpy, and pandas. (also sklearn for example datasets).
Prizes we're going for:
- Best Software Hack
- Best AI/ML Hack
- Best Beginner Software Hack
- Best Space App
Prizes Won
Team Members
Nikita Singh
Miranda Zhang
Kyler Schulz