What are the differences between LightGBM and FastTree algorithms used by ML.Net's modelbuilder?

Question

For my master thesis, I am using ML.Net library with the model builder to build machine learning models for regression / to predict values. My field of study is in constructional mechanics, so I'm rather new to machine learning.

For most of my models, tree-based regression algorithms seem to be the best performing, and of these LightGBM and FastTree are the best performing algorithms.

I've tried to read up on LightGBM here: https://www.microsoft.com/en-us/research/publication/lightgbm-a-highly-efficient-gradient-boosting-decision-tree/ And FastTree here: https://learn.microsoft.com/en-us/dotnet/api/microsoft.ml.trainers.fasttree.fasttreeregressiontrainer?view=ml-dotnet

However, I struggle to distinguish what the difference between these algorithms is. Could someone explain what the difference between LightGBM and FastTree is?

Answer 1

Hello,

Thanks for reaching out to us! The right place for ML.NET related question is the GitHub forum here, and the Gitter community. Sorry for the inconveniences.

Regards,
Yutong