Back to Community
Is 3D plotting possible in Quantopian Research?

Does Quantopian support 3D graphing in any way? It would be incredibly useful for visualising aspects of a new strategy I am developing.

I usually use the mplot3d toolkit from matplotlib, but it doesn't seem to be a valid import for Quantopian Research... Following a bit of research, it doesn't seem like Seaborn supports 3D plotting. Is there any way of plotting in 3D within a Research notebook, without simply linking to an image?

Thanks!

4 responses

To the best of my knowledge, we currently do not support 3D plotting. However, it'd obviously be useful for visualization, and it's easy to implement support for 3D plotting libraries. I expect that such capacities will be eventually added.

In the meantime, you may have to make do with 2D plotting. There are hacks and workarounds to communicate the same level of information (e.g. using multiple 2D plots). Though I haven't looked into it, I would also consider using dimensionality reduction techniques to collapse 3D-data into two dimensions while maintaining their relationships.

That is a shame, but oh well. As you said, I'm sure it will be implemented at some stage...

Instead of using multiple 2D plots, I could have got away to some extent using ternary plots (despite the semantics being iffy). These can be hacked together using matplotlib.tri, but again that isn't supported in Quantopian. Instead, seaborn FacetGrids will have to do...

Thanks.

Any update from the Q team on this library? It really limits research data visualization not being able to plot 3D charts.

Thanks

I was wondering the same thing about using Bokeh in research. Bokeh was utilized in a video tutorial provided by Q, but doesn't work in the actual research platform. For that reason I continue my research in a separate ipython notebook. Would love to be able to start using the research platform ....

https://www.quantopian.com/posts/bokeh-module-in-research