Skip to content

glassbox.cleaner.scalers

StandardScaler and MinMaxScaler for feature normalization.

StandardScaler 

StandardScaler()

Bases: BaseTransformer

Standardizes features by removing the mean and scaling to unit variance.

Source code in glassbox/cleaner/scalers.py 
16
17
18
def __init__(self):
    self._mean: np.ndarray = np.array([])
    self._std: np.ndarray = np.array([])

fit 

fit(X)

Compute the mean and standard deviation to be used for later scaling.

Parameters:

Name Type Description Default
X ndarray

Input array of shape (n_samples, n_features).

 required

Returns:

Type Description
Self

Fitted scaler instance.
Source code in glassbox/cleaner/scalers.py 
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
def fit(self, X: np.ndarray) -> Self:
    """
    Compute the mean and standard deviation to be used for later scaling.

    Parameters
    ----------
    X : np.ndarray
        Input array of shape (n_samples, n_features).

    Returns
    -------
    Self
        Fitted scaler instance.
    """
    n_features = X.shape[1]
    means = []
    stds = []

    for col_idx in range(n_features):
        col = X[:, col_idx]
        if np.issubdtype(col.dtype, np.number):
            col_clean = col[~np.isnan(col)]
        else:
            col_clean = np.array(
                [
                    x
                    for x in col
                    if x is not None and not (isinstance(x, float) and np.isnan(x))
                ]
            )

        if len(col_clean) == 0:
            means.append(0.0)
            stds.append(1.0)
        else:
            means.append(calc_mean(col_clean))
            s = calc_std(col_clean)
            stds.append(s if s > 0 else 1.0)

    self._mean = np.array(means)
    self._std = np.array(stds)
    return self

transform 

transform(X)

Perform standardization by centering and scaling.

Parameters:

Name Type Description Default
X ndarray

Input array of shape (n_samples, n_features).

 required

Returns:

Type Description
ndarray

Transformed array properly scaled.
Source code in glassbox/cleaner/scalers.py 
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
def transform(self, X: np.ndarray) -> np.ndarray:
    """
    Perform standardization by centering and scaling.

    Parameters
    ----------
    X : np.ndarray
        Input array of shape (n_samples, n_features).

    Returns
    -------
    np.ndarray
        Transformed array properly scaled.
    """
    X_out = X.copy()

    # If input has NaNs, they remain NaNs. We must be careful not to subtract mean from None
    if np.issubdtype(X_out.dtype, np.number):
        # broadcast subtraction
        X_out = (X_out - self._mean) / self._std
    else:
        for col_idx in range(X_out.shape[1]):
            col = X_out[:, col_idx]
            mask = np.array([v is not None for v in col])
            X_out[mask, col_idx] = (col[mask] - self._mean[col_idx]) / self._std[
                col_idx
            ]

    return X_out

MinMaxScaler 

MinMaxScaler()

Bases: BaseTransformer

Transforms features by scaling each feature to a given range.

Source code in glassbox/cleaner/scalers.py 
101
102
103
def __init__(self):
    self._min: np.ndarray = np.array([])
    self._max: np.ndarray = np.array([])

fit 

fit(X)

Compute the minimum and maximum to be used for later scaling.

Parameters:

Name Type Description Default
X ndarray

Input array of shape (n_samples, n_features).

 required

Returns:

Type Description
Self

Fitted scaler instance.
Source code in glassbox/cleaner/scalers.py 
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
def fit(self, X: np.ndarray) -> Self:
    """
    Compute the minimum and maximum to be used for later scaling.

    Parameters
    ----------
    X : np.ndarray
        Input array of shape (n_samples, n_features).

    Returns
    -------
    Self
        Fitted scaler instance.
    """
    n_features = X.shape[1]
    mins = []
    maxs = []

    for col_idx in range(n_features):
        col = X[:, col_idx]
        if np.issubdtype(col.dtype, np.number):
            col_clean = col[~np.isnan(col)]
        else:
            col_clean = np.array(
                [
                    x
                    for x in col
                    if x is not None and not (isinstance(x, float) and np.isnan(x))
                ]
            )

        if len(col_clean) == 0:
            mins.append(0.0)
            maxs.append(1.0)
        else:
            mins.append(float(np.min(col_clean)))
            maxs.append(float(np.max(col_clean)))

    self._min = np.array(mins)
    self._max = np.array(maxs)
    return self

transform 

transform(X)

Scale features of X according to feature range.

Parameters:

Name Type Description Default
X ndarray

Input array of shape (n_samples, n_features).

 required

Returns:

Type Description
ndarray

Transformed array properly scaled.
Source code in glassbox/cleaner/scalers.py 
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
def transform(self, X: np.ndarray) -> np.ndarray:
    """
    Scale features of X according to feature range.

    Parameters
    ----------
    X : np.ndarray
        Input array of shape (n_samples, n_features).

    Returns
    -------
    np.ndarray
        Transformed array properly scaled.
    """
    X_out = X.copy()
    range_vals = self._max - self._min
    range_vals[range_vals == 0.0] = 1.0  # avoid division by zero

    if np.issubdtype(X_out.dtype, np.number):
        X_out = (X_out - self._min) / range_vals
    else:
        for col_idx in range(X_out.shape[1]):
            col = X_out[:, col_idx]
            mask = np.array([v is not None for v in col])
            X_out[mask, col_idx] = (col[mask] - self._min[col_idx]) / range_vals[
                col_idx
            ]

    return X_out