IA Zaharia GeorgeEduard TSW Presentation

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Support Vector

Machines
George-Eduard Zaharia

Slow down and speed-up once in a while


Introduction - SVMs

◎ Used in a large array of domains (bioinformatics,


security)
◎ Supervised machine learning algorithm
◎ Used for classification or regression problems
◎ Separates data based on defined labels/outputs
◎ Introduced in 1963 and further developed in 1992
Classifying Data

◎ Find a separator between the


two groups

◎ Classify any new added point


Classifying Data
Maximum Margin
Model

◎ The points: (x1, y1), ... , (xn, yn)

◎ The hyperplane: f(X) = <w · X> + b

◎ The margin: 2/||w||

◎ Our purpose is to maximize the


margin
Soft-Margin SVM

◎ The hard margin SVM can work only when data is


completely linearly separable, without any errors

◎ Soft margin SVM allows error in the training set

◎ It produces a more general model when applied to


new datasets

Non-Linear Separation
Non-Linear Separation

Wherever we draw a line, the two groups of


points cannot be linearly separable
Non-Linear Separation

Mapping function: φ : X -> F


Non-Linear Separation
Kernel Trick

◎ Operate in a feature space without computing the


coordinates of the data in that space

◎ They compute the inner products between the images


of all pairs of data in the feature space

◎ Computationally cheaper
Kernel Example - Gaussian RBF Kernel
Kernel Example - Gaussian RBF Kernel
Other Kernel Functions

◎ Polynomial kernel: 𝐾(𝑋, 𝑌) = (𝑎 ∗ < 𝑋 · 𝑌 > + 𝑏) 𝑝

◎ Sigmoid kernel: 𝐾(𝑋, 𝑌) = 𝑡𝑎𝑛ℎ (𝑎 ∗ < 𝑋 · 𝑌 > + 𝑏)

◎ Linear kernel: 𝐾(𝑋, 𝑌) = < 𝑋 · 𝑌 > + 𝑏


Conclusion

Support vector machines prove to be strong classifiers


for various task:
● text categorization
● biology and other sciences
● image classification
THANK YOU!

You might also like