際際滷

際際滷Share a Scribd company logo

Machine Learning Wars: Deep Learning vs Gradient Boosting [Webinar]

Sefik Ilkin Serengil
Sefik Ilkin Serengil

This document discusses and compares gradient boosted machines (GBM) and deep learning. It notes that while deep learning can give powerful abilities, GBM may better explain models through decision trees. GBM works by combining many weak decision tree models to more accurately classify data, like moving a heavy rock bit by bit. The document cites GBM's success in Kaggle competitions, with it comprising the majority of winning solutions, sometimes combined with neural networks. In summary, the document presents GBM and deep learning, highlighting GBM's interpretability while also noting deep learning's strengths, and examples of GBM's competitive performance.

1 of 38
Download to read offline
Machine Learning Wars: Deep Learning vs GBM Sefik Ilkin Serengil
sefiks.com serengil Sefik Ilkin Serengil Software Developer 10 CS graduate, MSc @ 11 Instructor @
GBM Deep Learning
Deep Learning gives us superpowers
https://sefiks.com/2018/08/06/deep-face-recognition-with-keras/
https://sefiks.com/2018/01/01/facial-expression-recognition-with-keras/
https://sefiks.com/2018/07/20/artistic-style-transfer-with-deep-learning/
https://sefiks.com/2017/12/10/transfer-learning-in-keras-using-inception-v3/
Being strong is not enough sometimes
Machine Learning Wars: Deep Learning vs Gradient Boosting [Webinar]
But we cannot answer why and how questions as well
Machine Learning Wars: Deep Learning vs Gradient Boosting [Webinar]
GBM might slightly move the hammer
Machine Learning Wars: Deep Learning vs Gradient Boosting [Webinar]
https://archive.ics.uci.edu/ml/datasets/car+evaluation
Machine Learning Wars: Deep Learning vs Gradient Boosting [Webinar]
Machine Learning Wars: Deep Learning vs Gradient Boosting [Webinar]
700 lines of if statetements!
Machine Learning Wars: Deep Learning vs Gradient Boosting [Webinar]
Machine Learning Wars: Deep Learning vs Gradient Boosting [Webinar]
Machine Learning Wars: Deep Learning vs Gradient Boosting [Webinar]
Decision Trees are non-linear ML models
Classify this data set
Linear Model E.g. Perceptron, Linear SVM Decision Trees fails! They can handle
Now, classify this data set
Linear Model Decision Tree Simple linear model can easily classify this task Decision trees still can model the problem, but it wouldnt be easy!
GRADIENT BOOSTING
A single tree is not strong enough
100 m 70 m Illustration: How to explain Gradient Boosting - https://explained.ai/gradient-boosting/
30 m 15 m Illustration: How to explain Gradient Boosting - https://explained.ai/gradient-boosting/
15 m 15 m Illustration: How to explain Gradient Boosting - https://explained.ai/gradient-boosting/
Illustration: How to explain Gradient Boosting - https://explained.ai/gradient-boosting/
https://playground.tensorflow.org
http://arogozhnikov.github.io/2016/07/05/gradient_boosting_playground.html
Weak ones come together and move the heavy rock
XGBoost: A Scalable Tree Boosting System https://www.kdd.org/kdd2016/papers/files/rfp0697-chenAemb.pdf 2015 KAGGLE GRAND PRIX AMONG 29 WINNING SOLUTIONS Podium Ceremony 1 2 3 GBM 17 solutions Deep Neural Nets 11 solutions 9 solutions GBM + Neural Nets
Machine Learning Wars: Deep Learning vs Gradient Boosting [Webinar]
Thank you for your attention! sefiks.com serengil

More Related Content

Machine Learning Wars: Deep Learning vs Gradient Boosting [Webinar]