neurodata
Speed and memory benchmarks for complete overfitting
Are there speed and memory benchmarks (e.g. compared to scikit-learn, obliquetree etc.) for overfitting a dataset completely (no limit on tree depth)? This is usually a fair comparison for speed and memory without having to care about the accuracy part.
Hi @ogencoglu thanks for the question.
No not yet, and there are definitely speed limitations for certain tree implementations in treeple when it compares to scikit-learn for a variety of technical reasons. We would love to run it tho!
What is obliquetree? I could not find that package online?
Here:
- https://obliquetree.readthedocs.io/en/latest/
- https://github.com/sametcopur/obliquetree
Oh interesting I have not seen that yet. Our oblique tree implementation is slightly different tho just glancing at the implementation and description.
Would indeed be great to benchmark across the community!