sklearnkernels

sklearnkernels: extend functionality SVM and ANN classification and regression implementation
with alternative kernels proposed by Belanche.

the kernel hyperparameter manage is similar to SVM kernel parameter of sklearn library and could be some
member in brackets of next list

• Gausian (mrbf)
• Canberra (can)
• Truncated (tru)
• Hyperbolic (hyperbolic)
• Triangular (triangle)
• Radial Basic(radial_basic)
• Rational Quadratic(rquadratic)

Install

To install the library use pip:

pip install sklearnkernels

or clone the repo and just type the following on your shell:

python setup.py install

Usage examples

Example of usage:

from sklearnkernels.KSVM import KSVC, KSVR
from sklearnkernels.KANN import KANNC,KANNR

from sklearn import  datasets
from sklearn.model_selection import train_test_split
iris = datasets.load_iris()
Xc = iris.data
yc = iris.target
X_fc, X_tc, y_fc, y_tc = train_test_split(Xc, yc, test_size=0.25, random_state=0)
svc=KSVC(kernel='tru', C=100,gamma=0.001,degree=2)
svc.fit(X_fc, y_fc)
print('cmlfc',svc.score(X_tc,y_tc))

cmlfc 0.9210526315789473

Xr, yr = datasets.load_boston(return_X_y=True)
X_fr, X_tr, y_fr, y_tr = train_test_split(Xr, yr, test_size=0.25, random_state=0)
svr=KSVR(kernel='can', C=10,  gamma=300)
svr.fit(X_fr, y_fr)
print('clflr',svr.score(X_tr,y_tr))

clflr 0.7079035547838508

annc=KANNC(kernel='rbf',gamma=.25,random_state=0)
annc.fit(X_fc, y_fc)
print('clfk',annc.score(X_tc,y_tc)) 

clfk 0.8947368421052632

annr=KANNR(kernel='linear', gamma=1, early_stopping=True, max_iter=5000, random_state=0)
annr.fit(X_fr, y_fr)
print('clfk',annr.score(X_tr,y_tr))

clfk 0.5629257343580683