# Učitaj osnovne biblioteke...
import scipy as sp
import sklearn
%pylab inline

X = sp.array([[0],[1],[2],[4]])
y = sp.array([4,1,2,5])

from sklearn.preprocessing import PolynomialFeatures

from scipy import linalg

from sklearn.metrics import mean_squared_error

from sklearn.linear_model import LinearRegression

#generira N jednodimenzijskih primjera uniformno u intervalu [x1,x2]
def make_instances(x1,x2,N) :
    return sp.array([sp.array([x]) for x in linspace(x1,x2,N)])

from sklearn import cross_validation

from sklearn.linear_model import Ridge

def nonzeroes(coef) : return size(coef[coef!=0])

from sklearn.linear_model import Lasso

from sklearn.datasets import load_boston
boston = load_boston()
print boston.data.shape
print boston.target.shape

print boston.DESCR

from sklearn import cross_validation
X_train, X_rest, y_train, y_rest = cross_validation.train_test_split(boston.data,boston.target,train_size=0.6,random_state=42)
X_validate, X_test, y_validate, y_test = cross_validation.train_test_split(X_rest,y_rest,test_size=0.5,random_state=42)
print X_train.shape, X_validate.shape ,X_test.shape

seven_X = sp.array([[2,1],[2,3],[1,2],[3,2],[5,2],[5,4],[6,3]])
seven_y = sp.array([1,1,1,1,-1,-1,-1])

def plot_problem(X, y, h=None, surfaces=True) :
    '''
    Plots a two-dimensional labeled dataset (X,y) and, if function h(x) is given, 
    the decision boundaries (surfaces=False) or decision surfaces (surfaces=True)
    '''
    assert X.shape[1] == 2, "Dataset is not two-dimensional"
    if h!=None : 
        # Create a mesh to plot in
        r = 0.02  # mesh resolution
        x_min, x_max = X[:, 0].min() - 1, X[:, 0].max() + 1
        y_min, y_max = X[:, 1].min() - 1, X[:, 1].max() + 1
        xx, yy = np.meshgrid(np.arange(x_min, x_max, r),
                             np.arange(y_min, y_max, r))
        XX=np.c_[xx.ravel(), yy.ravel()]
        try:
            Z_test = h(XX)
            if shape(Z_test) == () :
                # h returns a scalar when applied to a matrix; map explicitly
                Z = sp.array(map(h,XX))
            else :
                Z = Z_test
        except ValueError:
            # can't apply to a matrix; map explicitly
            Z = sp.array(map(h,XX))
        # Put the result into a color plot
        Z = Z.reshape(xx.shape)
        if surfaces :
            plt.contourf(xx, yy, Z, cmap=plt.cm.Pastel1)
        else :
            plt.contour(xx, yy, Z)
    # Plot the dataset
    scatter(X[:,0],X[:,1],c=y, cmap=plt.cm.Paired,marker='o',s=50);

plot_problem(seven_X,seven_y)

from sklearn.linear_model import RidgeClassifier

X2 = sp.append(seven_X,[[2,2]],axis=0)
y2 = sp.append(seven_y,-1)

from sklearn.metrics import accuracy_score

X3 = sp.append(seven_X,[[12,8]],axis=0)
y3 = sp.append(seven_y,-1)

from sklearn.datasets import make_classification

from sklearn.linear_model import Perceptron

from sklearn.metrics import log_loss

from sklearn.metrics import zero_one_loss

from scipy.stats import multivariate_normal

from sklearn.linear_model import LogisticRegression

from sklearn.datasets import make_classification

from sklearn.preprocessing import PolynomialFeatures

#poly=PolynomialFeatures(2)
#...
#plot_problem(X,y,lambda x : model.predict(poly.transform(x))

from sklearn.cross_validation import train_test_split

data = sp.loadtxt("/path/to/glass.data", delimiter=",")
glass_X, glass_y = data[:,1:10], data[:,10]

from sklearn import cross_validation
X_train, X_test, y_train, y_test = cross_validation.train_test_split(glass_X,glass_y,train_size=2.0/3,random_state=42)
print X_train.shape, X_test.shape

from sklearn.preprocessing import StandardScaler