INTRODUCTION TO Support vector machines USING PYTHON USING THOR‘s HAMMER Miguel Cabrera @mfcabrera http://mfcabrera.com

OVERVIEW Tools Intro to Machine Learning Intuition behind SVM Basic usage Some situations Examples

TOOLS Every hero needs weapons

No content

No content

•  Powerful interactive shells (terminal and Qt-based). •  A browser-based notebook with support for code, text, mathematical expressions, inline plots and other rich media. •  Support for interactive data visualization and use of GUI toolkits. •  Flexible, embeddable interpreters to load into your own projects. •  Easy to use, high performance tools for parallel computing. From http://ipython.org :

Machine LEARNING

Machine Learning Supervised Classification Regression Unsupervised Clustering Feature Learning Reinforcement Learning, Recommender Systems, etc. No reason for Natalie Portman to be

Supervised x1 x2

UNSUPERVISED x1 x2

labels labels Supervised Training Tex, Documents, Imgages Training Tex, Documents, Imgages Training Tex, Documents, Imgages, Sounds labels Machine Learning Algorithm Predictive Model Features Vectors New Text Document, Images, Sounds Feature Vector Expected Label Adapted from: https://speakerdeck.com/ogrisel/machine-learning-in-python-with-scikit-learn

Applications Source: https://speakerdeck.com/ogrisel/trends-in-machine-learning-2

Spam Classification

TOPIC Classification

Sentiment analysis

OBJECT CLASSIFICATION

Support vector machines

Support vector machines? •  Effective in high dimensional spaces. •  Still effective in cases where number of dimensions is greater than the number of samples. •  Uses a subset of training points in the decision function (called support vectors), so it is also memory efficient. •  Easy to use and train •  Versatile: different kernel functions can be specified for the decision function.

WHEN?

Source: http://peekaboo-vision.blogspot.de/2013/01/machine-learning-cheat-sheet-for-scikit.html

THOR LIKES HIS HAMMER....

FOR THE FOLLOWING SECTIONS SEE THE IPYTHON NOTEBOOK FOUND AT: http://bit.ly/1g7VEfG

INTUITION BEHIND SVM

KERNELS

MULTICLASS

UNBALANCED DATA

TRAINING AND CROSS VALIDATIOn

Conclusion •  (I)Python / SciPy / Numpy / Scikit-Learn are awesome :D! •  SVM is mature algorithm, straightforward to use and works well in most of the cases. •  Being able to use Kernels allows SVM to learn complex decision boundaries. •  Using LibSVM / LibLinear based libraries allow for reusing models across languages or at least prototyping in Python.

QUESTIONS?

THANK YOU