Repository files navigation

Support vector machines flexible framework
We solve the unconstrained primal SVM formulation
SVM & Softmax classifiers supported
NB:
-Softmax classifier refers to penalized and kernalized logistic regression
-Classical logistic regression can be obtained by setting cost very high & using a linear kernel
-Python implementation in a seperate repository

LS: regression classifier using penalized least squared loss
Softmax: Softmax classifier using cross entropy loss
SVM:svm classifier using quadratic hinge loss

BGD: gradient descent (batch)
NGD: Newton-Raphson optimization (batch)
CGD: conjugate gradient descent (batch)
SGD: stochastic gradient descent (under development)

gaussian: gaussian kernel
linear: linear kernel
poly:polynomial kernel
For any remarks please let me knowazzouz.marouen@gmail.com