labelpropagation.py

# -*- coding: utf-8 -*-

'''
Label Propagationを実装した
'''

import numpy as np
import scipy as sp
import sklearn.datasets
import sklearn.metrics as metrics
from sklearn.model_selection import train_test_split
from sklearn.decomposition import PCA
from sklearn.svm import SVC
from sklearn.preprocessing import StandardScaler, MinMaxScaler


class LabelPropagation():
    '''
    Reference:
    Zhu, Xiaojin, Ghahramani, Zoubin. Learning from Labeled and Unlabeled Data with Label Propagation. 2002.
    '''
    def __init__(self, sigma=0.22):
        self.sigma = sigma

    def fit(self, X, y):
        self.X = X
        self.y = y

    def predict(self, X):
        data = np.concatenate((self.X, X), axis=0)

        dist_vec = sp.spatial.distance.pdist(data)
        dist_mat = sp.spatial.distance.squareform(dist_vec)
        weight = np.exp(- (np.multiply(dist_mat, dist_mat) / (self.sigma**2)))
        row_sum = np.sum(weight, axis=1)
        T = weight / row_sum
        print(T)

        self.dist = dist_mat
        self.weight = weight
        self.transition = T

        C = len(np.unique(self.y))
        print(C)
        Y_L = np.eye(C)[self.y]
        Y_U = np.zeros((X.shape[0], C))
        Y = np.concatenate((Y_L, Y_U), axis=0)

        Y_hat = Y
        for i in range(10):
            Y_hat = np.dot(T, Y_hat)

        u = np.argmax(Y_hat, axis=1)
        return u[self.y.shape[0]:]

    def score(self, X, y):
        pass



if __name__ == '__main__':
    data = sklearn.datasets.load_iris()
    #data = sklearn.datasets.load_digits()
    print(data.data.shape)

    X = data.data
    y = data.target
    X_train, X_test, y_train, y_test = train_test_split(
        X, y, test_size=0.25, random_state=52)

    # preprocess
    #transformer = StandardScaler()
    transformer = MinMaxScaler()
    X_train = transformer.fit_transform(X_train)
    X_test = transformer.transform(X_test)


    print('SVM')
    clf = SVC()
    clf.fit(X_train, y_train)
    predicted = clf.predict(X_test)
    print(metrics.accuracy_score(y_test, predicted))

    print('LP')
    clf = LabelPropagation()
    clf.fit(X_train, y_train)
    predicted = clf.predict(X_test)
    np.set_printoptions(threshold=np.inf)
    print(predicted)
    print(predicted.shape)
    print(metrics.accuracy_score(y_test, predicted))