#! /usr/bin/env python
# -*- coding: utf-8 -*-
"""
Extreme Learning Machine
This script is ELM for binary and multiclass classification.
"""
import numpy as np
from sklearn.base import BaseEstimator, ClassifierMixin
class ELM (BaseEstimator, ClassifierMixin):
"""
3 step model ELM
"""
def __init__(self,
hid_num,
a=1):
"""
Args:
hid_num (int): number of hidden neurons
a (int) : const value of sigmoid funcion
"""
self.hid_num = hid_num
self.a = a
def _sigmoid(self, x):
"""
sigmoid function
Args:
x (float): input
Returns:
float: output of sigmoid
"""
sigmoid_range = 34.538776394910684
x = np.clip(x, -sigmoid_range, sigmoid_range)
return 1 / (1 + np.exp(-self.a * x))
def _add_bias(self, X):
"""add bias to list
Args:
x_vs [[float]] Array: vec to add bias
Returns:
[float]: added vec
Examples:
>>> e = ELM(10, 3)
>>> e._add_bias(np.array([[1,2,3], [1,2,3]]))
array([[1., 2., 3., 1.],
[1., 2., 3., 1.]])
"""
return np.c_[X, np.ones(X.shape[0])]
def _ltov(self, n, label):
"""
trasform label scalar to vector
Args:
n (int) : number of class, number of out layer neuron
label (int) : label
Exmples:
>>> e = ELM(10, 3)
>>> e._ltov(3, 1)
[1, -1, -1]
>>> e._ltov(3, 2)
[-1, 1, -1]
>>> e._ltov(3, 3)
[-1, -1, 1]
"""
return [-1 if i != label else 1 for i in range(1, n + 1)]
def fit(self, X, y):
"""
learning
Args:
X [[float]] array : feature vectors of learnig data
y [[float]] array : labels of leanig data
"""
# number of class, number of output neuron
self.out_num = max(y)
if self.out_num != 1:
y = np.array([self._ltov(self.out_num, _y) for _y in y])
# add bias to feature vectors
X = self._add_bias(X)
# generate weights between input layer and hidden layer
np.random.seed()
self.W = np.random.uniform(-1., 1.,
(self.hid_num, X.shape[1]))
# find inverse weight matrix
_H = np.linalg.pinv(self._sigmoid(np.dot(self.W, X.T)))
self.beta = np.dot(_H.T, y)
return self
def predict(self, X):
"""
predict classify result
Args:
X [[float]] array: feature vectors of learnig data
Returns:
[int]: labels of classification result
"""
_H = self._sigmoid(np.dot(self.W, self._add_bias(X).T))
y = np.dot(_H.T, self.beta)
if self.out_num == 1:
return np.sign(y)
else:
return np.argmax(y, 1) + np.ones(y.shape[0])
def main():
from sklearn import preprocessing
from sklearn.datasets import fetch_openml as fetch_mldata
from sklearn.model_selection import ShuffleSplit, KFold, cross_val_score
db_name = 'australian'
hid_nums = [100, 200, 300]
data_set = fetch_mldata(db_name)
data_set.data = preprocessing.normalize(data_set.data)
data_set.target = [1 if i == 1 else -1
for i in data_set.target.astype(int)]
for hid_num in hid_nums:
print(hid_num, end=' ')
e = ELM(hid_num)
ave = 0
for i in range(10):
cv = KFold(n_splits=5, shuffle=True)
scores = cross_val_score(
e, data_set.data, data_set.target,
cv=cv, scoring='accuracy', n_jobs=-1)
ave += scores.mean()
ave /= 10
print("Accuracy: %0.3f " % (ave))
if __name__ == "__main__":
import doctest
doctest.testmod()
main()
