p-geon/normalize_by_sklearn.py

## normalize_by_sklearn.py
import copy
import itertools
from typing import List, Callable
import numpy as np; np.random.seed(seed=42)
from sklearn.preprocessing import StandardScaler, minmax_scale
# https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html#sklearn.preprocessing.StandardScaler
# https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.minmax_scale.html#sklearn.preprocessing.minmax_scale

def check_result(y1, y2, eps=1e-11):
    print(f'[y1] max: {np.round(np.max(y1), 4)}, min: {np.round(np.min(y1), 4)}, mean: {np.round(np.mean(y1), 4)}, std: {np.round(np.std(y1), 4)}')
    print(f'[y2] max: {np.round(np.max(y2), 4)}, min: {np.round(np.min(y2), 4)}, mean: {np.round(np.mean(y2), 4)}, std: {np.round(np.std(y2), 4)}')
    assert np.sum(np.linalg.norm(y1 - y2, ord=2)) < eps, f'Arrays inconsistency in normalizing. {y1}, {y2}'


def test_preprocessing(normalize: List[Callable], _y: np.ndarray) -> None:
    y = copy.deepcopy(_y)
    y1 = normalize[0](y)
    y2 = normalize[1](y)
    check_result(y1, y2)


ys = [np.identity(5)] + [np.random.rand(10, 10) for _ in range(5)]
methods = {
    'normalize_std': [
        lambda x: StandardScaler().fit(x).transform(x),
        lambda x: (x - np.mean(x, axis=0)) / np.std(x, axis=0)
     ],
    'normalize_minmax':  [
        lambda x: minmax_scale(x),
        lambda x: (x - np.min(x, axis=0))/(np.max(x, axis=0) - np.min(x, axis=0))
    ],
}


for i, (m, y) in enumerate(itertools.product(methods.keys(), ys)):
    print(f"> {m}-{1+i%ys.__len__()}")
    test_preprocessing(methods[m], y)
	import copy
	import itertools
	from typing import List, Callable
	import numpy as np; np.random.seed(seed=42)
	from sklearn.preprocessing import StandardScaler, minmax_scale
	# https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html#sklearn.preprocessing.StandardScaler
	# https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.minmax_scale.html#sklearn.preprocessing.minmax_scale

	def check_result(y1, y2, eps=1e-11):
	print(f'[y1] max: {np.round(np.max(y1), 4)}, min: {np.round(np.min(y1), 4)}, mean: {np.round(np.mean(y1), 4)}, std: {np.round(np.std(y1), 4)}')
	print(f'[y2] max: {np.round(np.max(y2), 4)}, min: {np.round(np.min(y2), 4)}, mean: {np.round(np.mean(y2), 4)}, std: {np.round(np.std(y2), 4)}')
	assert np.sum(np.linalg.norm(y1 - y2, ord=2)) < eps, f'Arrays inconsistency in normalizing. {y1}, {y2}'


	def test_preprocessing(normalize: List[Callable], _y: np.ndarray) -> None:
	y = copy.deepcopy(_y)
	y1 = normalize[0](y)
	y2 = normalize[1](y)
	check_result(y1, y2)


	ys = [np.identity(5)] + [np.random.rand(10, 10) for _ in range(5)]
	methods = {
	'normalize_std': [
	lambda x: StandardScaler().fit(x).transform(x),
	lambda x: (x - np.mean(x, axis=0)) / np.std(x, axis=0)
	],
	'normalize_minmax': [
	lambda x: minmax_scale(x),
	lambda x: (x - np.min(x, axis=0))/(np.max(x, axis=0) - np.min(x, axis=0))
	],
	}


	for i, (m, y) in enumerate(itertools.product(methods.keys(), ys)):
	print(f"> {m}-{1+i%ys.__len__()}")
	test_preprocessing(methods[m], y)