A Practical Guide to Tree-Based Learning Algorithms
sadanand-singh.github.io
A Practical Guide to Tree-Based Learning Algorithms
1–10 of 22 posts
Re: A Practical Guide to Tree-Based Learning Algorithms
#2I've found Adele Cutler's presentation on random forests to be an outstanding resource for getting intuition of tree-based algorithms.
http://www.math.usu.edu/adele/RandomForests/UofU2013.pdf
Thinking about trees as a supervised recursive partitioning algorithm or a clustering algorithm is useful for problems that may not appear to be simple classification or regression problems.
Re: A Practical Guide to Tree-Based Learning Algorithms
#3As interesting as I find the current state of deep learning to be, there is something about random forests that I can't help but find much more cool. Probably the amazing out-of-box performance.
Re: A Practical Guide to Tree-Based Learning Algorithms
#4As interesting as I find the current state of deep learning to be, there is something about random forests that I can't help but find much more cool. Probably the amazing out-of-box performance.
Yes, they have very few knobs to turn, which is very attractive.
Re: A Practical Guide to Tree-Based Learning Algorithms
#5Nice write-up, thanks for sharing. One possible typo I noticed:
> Maximum depth of tree (vertical depth) The maximum depth of trees. It is used to control over-fitting, higher values prevent a model from learning relations which might be highly specific to the particular sample.
Shouldn't it be lower values, i.e., shallower trees, that control over-fitting?
Re: A Practical Guide to Tree-Based Learning Algorithms
#6Nice write-up, thanks for sharing. One possible typo I noticed: > Maximum depth of tree (vertical depth) The maximum depth of trees. It is used to control over-fitting, higher values prevent a model from learning relations which might be highly specific to the particular sample. Shouldn't it be lower values , i.e., shallower trees, that control over-fitting?
Thanks for pointing. Yes it should be lower value to prevent over-fitting.
Re: A Practical Guide to Tree-Based Learning Algorithms
#7I've found Adele Cutler's presentation on random forests to be an outstanding resource for getting intuition of tree-based algorithms. http://www.math.usu.edu/adele/RandomForests/UofU2013.pdf Thinking about trees as a supervised recursive partitioning algorithm or a clustering algorithm is useful for problems that may not appear to be simple classification or regression problems.
On the topic of complementary resources, I really like Ben Gorman's explanation: https://gormanalysis.com/random-forest-from-top-to-bottom/. His related posts on singular decisions trees and GBM's are just as good, too.
Re: A Practical Guide to Tree-Based Learning Algorithms
#8I've found Adele Cutler's presentation on random forests to be an outstanding resource for getting intuition of tree-based algorithms. http://www.math.usu.edu/adele/RandomForests/UofU2013.pdf Thinking about trees as a supervised recursive partitioning algorithm or a clustering algorithm is useful for problems that may not appear to be simple classification or regression problems.
I like this one better:
https://web.csulb.edu/~tebert/teaching/lectures/551/random_f...
I made it.
Re: A Practical Guide to Tree-Based Learning Algorithms
#9Nice write-up, thanks for sharing. One possible typo I noticed: > Maximum depth of tree (vertical depth) The maximum depth of trees. It is used to control over-fitting, higher values prevent a model from learning relations which might be highly specific to the particular sample. Shouldn't it be lower values , i.e., shallower trees, that control over-fitting?
Thanks for pointing. Yes it should be lower value to prevent over-fitting.
df_train_set.Income.value_counts() should be df_test_set probably in the part where it's comparing both.
Re: A Practical Guide to Tree-Based Learning Algorithms
#10As interesting as I find the current state of deep learning to be, there is something about random forests that I can't help but find much more cool. Probably the amazing out-of-box performance.
also the model is analyzable so as to determine the variables which are contributing the most.