Ce diaporama a bien été signalé.
Nous utilisons votre profil LinkedIn et vos données d’activité pour vous proposer des publicités personnalisées et pertinentes. Vous pouvez changer vos préférences de publicités à tout moment.

Gradient Boosting Machines (GBM): from Zero to Hero (with R and Python code)

60 vues

Publié le

Data Con LA 2020
Description
This talk will get you started with gradient boosting machines (GBM), a

very popular machine learning technique providing state-of-the-art

accuracy on numerous business prediction problems. After a quick intro

to machine learning and the GBM algorithm, I will show how easy it is to

train and then use GBMs in real-life business applications using some

the most popular open source implementations (xgboost, lightgbm and

h2o). We'll do all this in both R and Python with only a few lines of

code and this talk will be accessible for a wide audience (with limited

prior knowledge of machine learning). Finally, in the last part of the

talk I will provide plenty of references that can get you to the next

level. GBMs are a powerful technique to have in your machine learning

toolbox, because despite all the latest hype about deep learning (neural

nets) and AI, in fact GBMs usually outperform neural networks on

structured/tabular data most often encountered in business applications.

Speaker
Szilard Pafka, Epoch, Chief Scientist

Publié dans : Données & analyses
  • Soyez le premier à commenter

  • Soyez le premier à aimer ceci

Gradient Boosting Machines (GBM): from Zero to Hero (with R and Python code)

  1. 1. Gradient Boosting Machines (GBM): From Zero to Hero (with R and Python Code) Szilard Pafka, PhD Chief Scientist, Epoch Data Con LA (Online) Oct 2020
  2. 2. Disclaimer: I am not representing my employer (Epoch) in this talk I cannot confirm nor deny if Epoch is using any of the methods, tools, results etc. mentioned in this talk
  3. 3. Source: Andrew Ng
  4. 4. Source: Andrew Ng
  5. 5. Source: Andrew Ng
  6. 6. Source: https://twitter.com/iamdevloper/
  7. 7. ...
  8. 8. y = f(x1 ,x2 ,...,xn ) “Learn” f from data
  9. 9. y = f(x1 ,x2 ,...,xn )
  10. 10. y = f(x1 ,x2 ,...,xn )
  11. 11. Supervised Learning Data: X (n obs, p features), y (labels) Regression, classification Train/learn/fit f from data (model) Score: for new x, get f(x) Algos: LR, k-NN, DT, RF, GBM, NN/DL, SVM, NB… Goal: max acc/min err new data Metrics: MSE, AUC (ROC) Bad: measure on train set. Need: test set/cross-validation (CV) Hyperparameters, model capacity, overfitting Regularization Model selection Hyperparameter search (grid, random) Ensembles
  12. 12. Supervised Learning Data: X (n obs, p features), y (labels) Regression, classification Train/learn/fit f from data (model) Score: for new x, get f(x) Algos: LR, k-NN, DT, RF, GBM, NN/DL, SVM, NB… Goal: max acc/min err new data Metrics: MSE, AUC (ROC) Bad: measure on train set. Need: test set/cross-validation (CV) Hyperparameters, model capacity, overfitting Regularization Model selection Hyperparameter search (grid, random) Ensembles
  13. 13. Source: Hastie etal, ESL 2ed
  14. 14. Source: Hastie etal, ESL 2ed
  15. 15. no-one is using this crap
  16. 16. Live Demo Summary of the demo for those reading just the slides (e.g. those who did not attend the talk):
  17. 17. http://www.jmlr.org/papers/volume13/bergstra12a/bergstra12a.pdf
  18. 18. End of Demo

×