Additionally, the small amount of research that is done on GBMs is often paywalled, not open sourced or released and never maintained. It would be great if academia gave extended grants to open source and maintain projects for X years after publication.
Very enlightening. Will this be followed up by some practical examples of application of boosted trees to real world data? I work in aerospace manufacturing and having spent several years building up a strong foundation for realiable data now we have started progressing from the descriptive to the diagnostic and I'm going to start applying boosted trees to understand feature importance.
Additionally, the small amount of research that is done on GBMs is often paywalled, not open sourced or released and never maintained. It would be great if academia gave extended grants to open source and maintain projects for X years after publication.
I like your series of posts! As a quick note, the first sentence under title "We Need More GBT Research" says:
"""
Even though GBTs are extremely capable and successful, there has been relatively little research on the.
"""
I guess the last word should be "them" and not "the". :)
Hum, I started to rewrite a code in c++ but now i'm without time to continue
Maybe this lib should be interesting the ideas are very nice, it's R based, not python: https://github.com/OVVO-Financial/NNS
example with xgb - https://github.com/OVVO-Financial/NNS/blob/NNS-Beta-Version/examples/xgboost_example.html (no idea how to open the html parsed file)
Very enlightening. Will this be followed up by some practical examples of application of boosted trees to real world data? I work in aerospace manufacturing and having spent several years building up a strong foundation for realiable data now we have started progressing from the descriptive to the diagnostic and I'm going to start applying boosted trees to understand feature importance.
I would love that as well