Spaces:
Running
Running
File size: 2,477 Bytes
bdf9096 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 |
import numpy as np
from sklearn.base import TransformerMixin, BaseEstimator
from utils.param_keys import N_COMPONENTS
class MyModel(TransformerMixin, BaseEstimator):
"""
This class and some child classes are partly copied from:
https://towardsdatascience.com/implementing-pca-from-scratch-fb434f1acbaa
and commented with the help of:
https://www.askpython.com/python/examples/principal-component-analysis
"""
def __init__(self):
self.explained_variance_ = None
self.components_ = None
self._standardized_data = None
self.n_components = None
self.n_samples = None
self._covariance_matrix = None
def __str__(self):
return f'{self.__class__.__name__}:\ncomponents={self.n_components}'
def fit_transform(self, data_ndarray, **fit_params):
self.fit(data_ndarray, **fit_params)
return self.transform(data_ndarray)
def fit(self, data_matrix, **fit_params):
self.n_samples = data_matrix.shape[0]
self.n_components = fit_params.get(N_COMPONENTS, 2)
self._standardized_data = self._standardize_data(data_matrix)
self._covariance_matrix = self.get_covariance_matrix()
self.components_ = self.get_eigenvectors()
return self
@staticmethod
def _standardize_data(matrix):
"""
Subtract mean and divide by standard deviation column-wise.
Doing this proves to be very helpful when calculating the covariance matrix.
https://towardsdatascience.com/understanding-the-covariance-matrix-92076554ea44
Mean-Center the data
:param matrix: Data as matrix
:return: Standardized data matrix
"""
numerator = matrix - np.mean(matrix, axis=0)
denominator = np.std(matrix, axis=0)
return numerator / denominator
def get_covariance_matrix(self):
"""
Calculate covariance matrix with standardized matrix A
:return: Covariance Matrix
"""
return np.cov(self._standardized_data.T)
def get_eigenvectors(self):
pass
def transform(self, data_matrix):
"""
Project the data to the lower dimension with the help of the eigenvectors.
:return: Data reduced to lower dimensions from higher dimensions
"""
data_matrix_standardized = self._standardize_data(data_matrix)
return np.dot(data_matrix_standardized, self.components_[:, :self.n_components])
|