Skip to content

Parameters

Categorical

Bases: Parameter

Variable with a categorical domain (nominal scale, values cannot be put into order).

Attributes:

Name Type Description
name str

name of the parameter

domain list

list possible values

Source code in opti/parameter.py
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
class Categorical(Parameter):
    """Variable with a categorical domain (nominal scale, values cannot be put into order).

    Attributes:
        name (str): name of the parameter
        domain (list): list possible values
    """

    def __init__(self, name: str, domain: List[str], **kwargs):
        if not isinstance(domain, list):
            raise TypeError(f"{name}: Domain must be of type list.")
        if len(domain) < 2:
            raise ValueError(f"{name}: Domain must a least contain 2 values.")
        if len(set(domain)) != len(domain):
            raise ValueError(f"{name}: Domain contains duplicates.")
        super().__init__(name, domain, type="categorical", **kwargs)

    def __repr__(self):
        return f"Categorical('{self.name}', domain={self.domain})"

    @property
    def bounds(self) -> Tuple[float, float]:
        """Return the domain bounds."""
        return np.nan, np.nan

    def contains(self, point):
        """Check if a point is in contained in the domain.

        Args:
            point (float, np.ndarray, pd.Series or pd.Dataframe): parameter value(s).

        Returns:
            Object of the same type as `point` with boolean datatype.
        """
        if not np.isscalar(point):
            point = np.array(point)
        return np.isin(point, self.domain)

    def round(self, point):
        """Round a point to the closest contained values.

        Args:
            point (float, np.ndarray, pd.Series or pd.Dataframe): parameter value(s).

        Returns:
            Object of the same type as `point`.
        """
        if not np.all(self.contains(point)):
            raise ValueError(
                f"{self.name}: Cannot round values for categorical parameter."
            )
        return point

    def sample(self, n: int = 1) -> pd.Series:
        """Draw random samples from the domain."""
        return pd.Series(name=self.name, data=np.random.choice(self.domain, n))

    def to_onehot_encoding(self, points: pd.Series) -> pd.DataFrame:
        """Convert points to a one-hot encoding."""
        return pd.DataFrame(
            {f"{self.name}{_CAT_SEP}{c}": points == c for c in self.domain}, dtype=float
        )

    def from_onehot_encoding(self, points: pd.DataFrame) -> pd.Series:
        """Convert points back from one-hot encoding."""
        cat_cols = [f"{self.name}{_CAT_SEP}{c}" for c in self.domain]
        if np.any([c not in cat_cols for c in points.columns]):
            raise ValueError(
                f"{self.name}: Column names don't match categorical levels: {points.columns}, {cat_cols}."
            )
        s = points.idxmax(1).str.split(_CAT_SEP, expand=True)[1]
        s.name = self.name
        return s

    def to_dummy_encoding(self, points: pd.Series) -> pd.DataFrame:
        """Convert points to a dummy-hot encoding, dropping the first categorical level."""
        return pd.DataFrame(
            {f"{self.name}{_CAT_SEP}{c}": points == c for c in self.domain[1:]},
            dtype=float,
        )

    def from_dummy_encoding(self, points: pd.DataFrame) -> pd.Series:
        """Convert points back from dummy encoding."""
        cat_cols = [f"{self.name}{_CAT_SEP}{c}" for c in self.domain]
        if np.any([c not in cat_cols[1:] for c in points.columns]):
            raise ValueError(
                f"{self.name}: Column names don't match categorical levels: {points.columns}, {cat_cols}."
            )
        points = points.copy()
        points[cat_cols[0]] = 1 - points[cat_cols[1:]].sum(axis=1)
        s = points.idxmax(1).str.split(_CAT_SEP, expand=True)[1]
        s.name = self.name
        return s

    def to_label_encoding(self, points: pd.Series) -> pd.Series:
        """Convert points to label-encoding."""
        enc = pd.Series(range(len(self.domain)), index=list(self.domain))
        s = enc[points]
        s.index = points.index
        s.name = self.name
        return s

    def from_label_encoding(self, points: pd.Series) -> pd.Series:
        """Convert points back from label-encoding."""
        enc = np.array(self.domain)
        return pd.Series(enc[points], index=points.index)

bounds property

Return the domain bounds.

contains(point)

Check if a point is in contained in the domain.

Parameters:

Name Type Description Default
point (float, ndarray, Series or Dataframe)

parameter value(s).

required

Returns:

Type Description

Object of the same type as point with boolean datatype.

Source code in opti/parameter.py
248
249
250
251
252
253
254
255
256
257
258
259
def contains(self, point):
    """Check if a point is in contained in the domain.

    Args:
        point (float, np.ndarray, pd.Series or pd.Dataframe): parameter value(s).

    Returns:
        Object of the same type as `point` with boolean datatype.
    """
    if not np.isscalar(point):
        point = np.array(point)
    return np.isin(point, self.domain)

from_dummy_encoding(points)

Convert points back from dummy encoding.

Source code in opti/parameter.py
304
305
306
307
308
309
310
311
312
313
314
315
def from_dummy_encoding(self, points: pd.DataFrame) -> pd.Series:
    """Convert points back from dummy encoding."""
    cat_cols = [f"{self.name}{_CAT_SEP}{c}" for c in self.domain]
    if np.any([c not in cat_cols[1:] for c in points.columns]):
        raise ValueError(
            f"{self.name}: Column names don't match categorical levels: {points.columns}, {cat_cols}."
        )
    points = points.copy()
    points[cat_cols[0]] = 1 - points[cat_cols[1:]].sum(axis=1)
    s = points.idxmax(1).str.split(_CAT_SEP, expand=True)[1]
    s.name = self.name
    return s

from_label_encoding(points)

Convert points back from label-encoding.

Source code in opti/parameter.py
325
326
327
328
def from_label_encoding(self, points: pd.Series) -> pd.Series:
    """Convert points back from label-encoding."""
    enc = np.array(self.domain)
    return pd.Series(enc[points], index=points.index)

from_onehot_encoding(points)

Convert points back from one-hot encoding.

Source code in opti/parameter.py
286
287
288
289
290
291
292
293
294
295
def from_onehot_encoding(self, points: pd.DataFrame) -> pd.Series:
    """Convert points back from one-hot encoding."""
    cat_cols = [f"{self.name}{_CAT_SEP}{c}" for c in self.domain]
    if np.any([c not in cat_cols for c in points.columns]):
        raise ValueError(
            f"{self.name}: Column names don't match categorical levels: {points.columns}, {cat_cols}."
        )
    s = points.idxmax(1).str.split(_CAT_SEP, expand=True)[1]
    s.name = self.name
    return s

round(point)

Round a point to the closest contained values.

Parameters:

Name Type Description Default
point (float, ndarray, Series or Dataframe)

parameter value(s).

required

Returns:

Type Description

Object of the same type as point.

Source code in opti/parameter.py
261
262
263
264
265
266
267
268
269
270
271
272
273
274
def round(self, point):
    """Round a point to the closest contained values.

    Args:
        point (float, np.ndarray, pd.Series or pd.Dataframe): parameter value(s).

    Returns:
        Object of the same type as `point`.
    """
    if not np.all(self.contains(point)):
        raise ValueError(
            f"{self.name}: Cannot round values for categorical parameter."
        )
    return point

sample(n=1)

Draw random samples from the domain.

Source code in opti/parameter.py
276
277
278
def sample(self, n: int = 1) -> pd.Series:
    """Draw random samples from the domain."""
    return pd.Series(name=self.name, data=np.random.choice(self.domain, n))

to_dummy_encoding(points)

Convert points to a dummy-hot encoding, dropping the first categorical level.

Source code in opti/parameter.py
297
298
299
300
301
302
def to_dummy_encoding(self, points: pd.Series) -> pd.DataFrame:
    """Convert points to a dummy-hot encoding, dropping the first categorical level."""
    return pd.DataFrame(
        {f"{self.name}{_CAT_SEP}{c}": points == c for c in self.domain[1:]},
        dtype=float,
    )

to_label_encoding(points)

Convert points to label-encoding.

Source code in opti/parameter.py
317
318
319
320
321
322
323
def to_label_encoding(self, points: pd.Series) -> pd.Series:
    """Convert points to label-encoding."""
    enc = pd.Series(range(len(self.domain)), index=list(self.domain))
    s = enc[points]
    s.index = points.index
    s.name = self.name
    return s

to_onehot_encoding(points)

Convert points to a one-hot encoding.

Source code in opti/parameter.py
280
281
282
283
284
def to_onehot_encoding(self, points: pd.Series) -> pd.DataFrame:
    """Convert points to a one-hot encoding."""
    return pd.DataFrame(
        {f"{self.name}{_CAT_SEP}{c}": points == c for c in self.domain}, dtype=float
    )

Continuous

Bases: Parameter

Variable that can take on any real value in the specified domain.

Attributes:

Name Type Description
name str

name of the parameter

domain list

[lower bound, upper bound]

Source code in opti/parameter.py
 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
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
class Continuous(Parameter):
    """Variable that can take on any real value in the specified domain.

    Attributes:
        name (str): name of the parameter
        domain (list): [lower bound, upper bound]
    """

    def __init__(
        self,
        name: str,
        domain: Optional[Sequence] = None,
        **kwargs,
    ):
        if domain is None:
            domain = [-np.inf, np.inf]
        else:
            if len(domain) != 2:
                raise ValueError(
                    f"{name}: Domain must consist of two values [low, high]."
                )
        # convert None to +/- inf and string to float
        low = -np.inf if domain[0] is None else float(domain[0])
        high = np.inf if domain[1] is None else float(domain[1])

        if high < low:
            raise ValueError(
                f"{name}: Lower bound {low} must be less than upper bound {high}."
            )
        self.low = low
        self.high = high
        super().__init__(name=name, domain=[low, high], type="continuous", **kwargs)

    def __repr__(self):
        if np.isfinite(self.low) or np.isfinite(self.high):
            return f"Continuous('{self.name}', domain={self.domain})"
        else:
            return f"Continuous('{self.name}')"

    @property
    def bounds(self) -> Tuple[float, float]:
        """Return the domain bounds."""
        return self.low, self.high

    def contains(self, point):
        """Check if a point is in contained in the domain.

        Args:
            point (float, np.ndarray, pd.Series or pd.Dataframe): parameter value(s).

        Returns:
            Object of the same type as `point` with boolean datatype.
        """
        return (self.low <= point) & (point <= self.high)

    def round(self, point):
        """Round a point to the closest contained values.

        Args:
            point (float, np.ndarray, pd.Series or pd.Dataframe): parameter value(s).

        Returns:
            Object of the same type as `point` with values clipped to parameter bounds.
        """
        return np.clip(point, self.low, self.high)

    def sample(self, n: int = 1) -> pd.Series:
        """Draw random samples from the domain."""
        low = max(self.low, np.finfo(np.float32).min)
        high = min(self.high, np.finfo(np.float32).max)
        return pd.Series(name=self.name, data=np.random.uniform(low, high, n))

    def to_config(self) -> dict:
        """Return a json-serializable configuration dict."""
        conf = dict(name=self.name, type=self.type)
        low = None if np.isinf(self.low) else float(self.low)
        high = None if np.isinf(self.high) else float(self.high)
        if low is not None or high is not None:
            conf.update({"domain": [low, high]})
        conf.update(self.extra_fields)
        return conf

    def to_unit_range(self, points):
        """Transform points to the unit range: [low, high] -> [0, 1].

        Points outside of the domain will transform to outside of [0, 1].
        Nothing is done if low == high.
        """
        if np.isclose(self.low, self.high):
            return points
        else:
            return (points - self.low) / (self.high - self.low)

    def from_unit_range(self, points):
        """Transform points from the unit range: [0, 1] -> [low, high].

        A rounding is applied to correct for numerical precision.
        Nothing is done if low == high.
        """
        if np.isclose(self.low, self.high):
            return points
        else:
            return points * (self.high - self.low) + self.low

bounds property

Return the domain bounds.

contains(point)

Check if a point is in contained in the domain.

Parameters:

Name Type Description Default
point (float, ndarray, Series or Dataframe)

parameter value(s).

required

Returns:

Type Description

Object of the same type as point with boolean datatype.

Source code in opti/parameter.py
70
71
72
73
74
75
76
77
78
79
def contains(self, point):
    """Check if a point is in contained in the domain.

    Args:
        point (float, np.ndarray, pd.Series or pd.Dataframe): parameter value(s).

    Returns:
        Object of the same type as `point` with boolean datatype.
    """
    return (self.low <= point) & (point <= self.high)

from_unit_range(points)

Transform points from the unit range: [0, 1] -> [low, high].

A rounding is applied to correct for numerical precision. Nothing is done if low == high.

Source code in opti/parameter.py
119
120
121
122
123
124
125
126
127
128
def from_unit_range(self, points):
    """Transform points from the unit range: [0, 1] -> [low, high].

    A rounding is applied to correct for numerical precision.
    Nothing is done if low == high.
    """
    if np.isclose(self.low, self.high):
        return points
    else:
        return points * (self.high - self.low) + self.low

round(point)

Round a point to the closest contained values.

Parameters:

Name Type Description Default
point (float, ndarray, Series or Dataframe)

parameter value(s).

required

Returns:

Type Description

Object of the same type as point with values clipped to parameter bounds.

Source code in opti/parameter.py
81
82
83
84
85
86
87
88
89
90
def round(self, point):
    """Round a point to the closest contained values.

    Args:
        point (float, np.ndarray, pd.Series or pd.Dataframe): parameter value(s).

    Returns:
        Object of the same type as `point` with values clipped to parameter bounds.
    """
    return np.clip(point, self.low, self.high)

sample(n=1)

Draw random samples from the domain.

Source code in opti/parameter.py
92
93
94
95
96
def sample(self, n: int = 1) -> pd.Series:
    """Draw random samples from the domain."""
    low = max(self.low, np.finfo(np.float32).min)
    high = min(self.high, np.finfo(np.float32).max)
    return pd.Series(name=self.name, data=np.random.uniform(low, high, n))

to_config()

Return a json-serializable configuration dict.

Source code in opti/parameter.py
 98
 99
100
101
102
103
104
105
106
def to_config(self) -> dict:
    """Return a json-serializable configuration dict."""
    conf = dict(name=self.name, type=self.type)
    low = None if np.isinf(self.low) else float(self.low)
    high = None if np.isinf(self.high) else float(self.high)
    if low is not None or high is not None:
        conf.update({"domain": [low, high]})
    conf.update(self.extra_fields)
    return conf

to_unit_range(points)

Transform points to the unit range: [low, high] -> [0, 1].

Points outside of the domain will transform to outside of [0, 1]. Nothing is done if low == high.

Source code in opti/parameter.py
108
109
110
111
112
113
114
115
116
117
def to_unit_range(self, points):
    """Transform points to the unit range: [low, high] -> [0, 1].

    Points outside of the domain will transform to outside of [0, 1].
    Nothing is done if low == high.
    """
    if np.isclose(self.low, self.high):
        return points
    else:
        return (points - self.low) / (self.high - self.low)

Discrete

Bases: Parameter

Variable with a discrete domain (ordinal scale).

Attributes:

Name Type Description
name str

name of the parameter

domain list

list of possible numeric values

Source code in opti/parameter.py
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
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
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
class Discrete(Parameter):
    """Variable with a discrete domain (ordinal scale).

    Attributes:
        name (str): name of the parameter
        domain (list): list of possible numeric values
    """

    def __init__(self, name: str, domain: Sequence, **kwargs):
        if len(domain) < 1:
            raise ValueError(f"{name}: Domain must contain at least one value.")
        try:
            # convert to a sorted list of floats
            domain = np.sort(np.array(domain).astype(float)).tolist()
        except ValueError:
            raise ValueError(f"{name}: Domain contains non-numeric values.")
        if len(set(domain)) != len(domain):
            raise ValueError(f"{name}: Domain contains duplicates.")
        self.low = min(domain)
        self.high = max(domain)
        super().__init__(name, domain, type="discrete", **kwargs)

    def __repr__(self):
        return f"Discrete('{self.name}', domain={self.domain})"

    @property
    def bounds(self) -> Tuple[float, float]:
        """Return the domain bounds."""
        return self.low, self.high

    def contains(self, point):
        """Check if a point is in contained in the domain.

        Args:
            point (float, np.ndarray, pd.Series or pd.Dataframe): parameter value(s).

        Returns:
            Object of the same type as `point` with boolean datatype.
        """
        if not np.isscalar(point):
            point = np.array(point)
        return np.isin(point, self.domain)

    def round(self, point):
        """Round a point to the closest contained values.

        Args:
            point (float, np.ndarray, pd.Series or pd.Dataframe): parameter value(s).

        Returns:
            Object of the same type as `point` with values clipped to parameter bounds.
        """
        if np.isscalar(point):
            i = np.argmin(np.abs(np.array(self.domain) - point))
            return self.domain[i]
        closest = [np.argmin(np.abs(np.array(self.domain) - p)) for p in point]
        rounded = np.array(self.domain)[closest]
        if isinstance(point, np.ndarray):
            return rounded
        elif isinstance(point, pd.Series):
            return pd.Series(name=self.name, data=rounded, index=point.index)
        elif isinstance(point, pd.DataFrame):
            return pd.DataFrame({self.name: rounded}, index=point.index)

    def sample(self, n: int = 1) -> pd.Series:
        """Draw random samples from the domain."""
        return pd.Series(name=self.name, data=np.random.choice(self.domain, n))

    def to_unit_range(self, points):
        """Transform points to the unit range: [low, high] -> [0, 1].

        Points outside of the domain will transform to outside of [0, 1].
        Nothing is done if low == high.
        """
        if np.isclose(self.low, self.high):
            return points
        else:
            return (points - self.low) / (self.high - self.low)

    def from_unit_range(self, points):
        """Transform points from the unit range: [0, 1] -> [low, high].

        A rounding is applied to correct for numerical precision.
        Nothing is done if low == high.
        """
        if np.isclose(self.low, self.high):
            return points
        else:
            points = points * (self.high - self.low) + self.low
            return self.round(points)

bounds property

Return the domain bounds.

contains(point)

Check if a point is in contained in the domain.

Parameters:

Name Type Description Default
point (float, ndarray, Series or Dataframe)

parameter value(s).

required

Returns:

Type Description

Object of the same type as point with boolean datatype.

Source code in opti/parameter.py
161
162
163
164
165
166
167
168
169
170
171
172
def contains(self, point):
    """Check if a point is in contained in the domain.

    Args:
        point (float, np.ndarray, pd.Series or pd.Dataframe): parameter value(s).

    Returns:
        Object of the same type as `point` with boolean datatype.
    """
    if not np.isscalar(point):
        point = np.array(point)
    return np.isin(point, self.domain)

from_unit_range(points)

Transform points from the unit range: [0, 1] -> [low, high].

A rounding is applied to correct for numerical precision. Nothing is done if low == high.

Source code in opti/parameter.py
210
211
212
213
214
215
216
217
218
219
220
def from_unit_range(self, points):
    """Transform points from the unit range: [0, 1] -> [low, high].

    A rounding is applied to correct for numerical precision.
    Nothing is done if low == high.
    """
    if np.isclose(self.low, self.high):
        return points
    else:
        points = points * (self.high - self.low) + self.low
        return self.round(points)

round(point)

Round a point to the closest contained values.

Parameters:

Name Type Description Default
point (float, ndarray, Series or Dataframe)

parameter value(s).

required

Returns:

Type Description

Object of the same type as point with values clipped to parameter bounds.

Source code in opti/parameter.py
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
def round(self, point):
    """Round a point to the closest contained values.

    Args:
        point (float, np.ndarray, pd.Series or pd.Dataframe): parameter value(s).

    Returns:
        Object of the same type as `point` with values clipped to parameter bounds.
    """
    if np.isscalar(point):
        i = np.argmin(np.abs(np.array(self.domain) - point))
        return self.domain[i]
    closest = [np.argmin(np.abs(np.array(self.domain) - p)) for p in point]
    rounded = np.array(self.domain)[closest]
    if isinstance(point, np.ndarray):
        return rounded
    elif isinstance(point, pd.Series):
        return pd.Series(name=self.name, data=rounded, index=point.index)
    elif isinstance(point, pd.DataFrame):
        return pd.DataFrame({self.name: rounded}, index=point.index)

sample(n=1)

Draw random samples from the domain.

Source code in opti/parameter.py
195
196
197
def sample(self, n: int = 1) -> pd.Series:
    """Draw random samples from the domain."""
    return pd.Series(name=self.name, data=np.random.choice(self.domain, n))

to_unit_range(points)

Transform points to the unit range: [low, high] -> [0, 1].

Points outside of the domain will transform to outside of [0, 1]. Nothing is done if low == high.

Source code in opti/parameter.py
199
200
201
202
203
204
205
206
207
208
def to_unit_range(self, points):
    """Transform points to the unit range: [low, high] -> [0, 1].

    Points outside of the domain will transform to outside of [0, 1].
    Nothing is done if low == high.
    """
    if np.isclose(self.low, self.high):
        return points
    else:
        return (points - self.low) / (self.high - self.low)

Parameter

Parameter base class.

Source code in opti/parameter.py
10
11
12
13
14
15
16
17
18
19
20
21
22
23
class Parameter:
    """Parameter base class."""

    def __init__(self, name: str, domain: Sequence, type: str = None, **kwargs):
        self.name = name
        self.domain = domain
        self.type = type
        self.extra_fields = kwargs

    def to_config(self) -> dict:
        """Return a json-serializable configuration dict."""
        conf = dict(name=self.name, type=self.type, domain=self.domain)
        conf.update(self.extra_fields)
        return conf

to_config()

Return a json-serializable configuration dict.

Source code in opti/parameter.py
19
20
21
22
23
def to_config(self) -> dict:
    """Return a json-serializable configuration dict."""
    conf = dict(name=self.name, type=self.type, domain=self.domain)
    conf.update(self.extra_fields)
    return conf

Parameters

Set of parameters representing either the input or the output parameter space.

Source code in opti/parameter.py
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
class Parameters:
    """Set of parameters representing either the input or the output parameter space."""

    def __init__(self, parameters):
        """
        It can be constructed either from a list / tuple (of at least one) Parameter objects
        ```
        Parameters([
            Continuous(name="foo", domain=[1, 10]),
            Discrete(name="bar", domain=[1, 2, 3, 4]),
            Categorical(name="baz", domain=["A", "B", 3]),
        ])
        ```
        or from a list / tuple of dicts
        ```
        Parameters([
            {"name": "foo", "type": "continuous", "domain": [1, 10]},
            {"name": "bar", "type": "discrete", "domain": [1, 2, 3, 4]},
            {"name": "baz", "type": "categorical", "domain": ["A", "B", 3]},
            {"name": "baz", "type": "categorical", "domain": ["A", "B", 3], extra="info"},
        ])
        ```
        In particular, Parameters().__init__ and the to_config method of each Parameter type are inverses.
        Parameters(conf).to_config() == conf
        """
        if not isinstance(parameters, (list, tuple)):
            raise TypeError("Parameters expects a list or tuple of parameters.")

        self.parameters = {}
        for d in parameters:
            if not isinstance(d, Parameter):
                d = make_parameter(**d)
            if d.name in self.parameters:
                raise ValueError(f"Duplicate parameter name {d.name}")
            self.parameters[d.name] = d

    def __repr__(self):
        return "Parameters(\n" + pprint.pformat(list(self.parameters.values())) + "\n)"

    def __iter__(self):
        return iter(self.parameters.values())

    def __getitem__(self, name):
        return self.parameters[name]

    def __len__(self):
        return len(self.parameters)

    def __add__(self, other):
        parameters = list(self.parameters.values()) + list(other.parameters.values())
        return Parameters(parameters)

    @property
    def names(self):
        return list(self.parameters.keys())

    @property
    def bounds(self) -> pd.DataFrame:
        """Return the parameter bounds."""
        return pd.DataFrame({p.name: p.bounds for p in self}, index=["min", "max"])

    def contains(self, points: pd.DataFrame) -> pd.Series:
        """Check if points are inside the domain of each parameter."""
        if isinstance(points, pd.DataFrame):
            points = points[self.names]
        b = np.stack([self[k].contains(v) for k, v in points.items()], axis=1)
        return b.all(axis=1)

    def round(self, points: pd.DataFrame) -> pd.DataFrame:
        """Round points to the closest contained values."""
        return pd.concat([self[k].round(v) for k, v in points.items()], axis=1)

    def sample(self, n: int = 1) -> pd.DataFrame:
        """Draw uniformly distributed random samples."""
        return pd.concat([param.sample(n) for param in self], axis=1)

    def transform(
        self,
        points: pd.DataFrame,
        continuous: str = "none",
        discrete: str = "none",
        categorical: str = "onehot-encode",
    ) -> pd.DataFrame:
        """Transfrom the given dataframe according to a set of transformation rules.

        Args:
            points (pd.DataFrame): Dataframe to transfrom. Must contain columns for each parameter and may contain additional columns.
            continuous (str, optional): Transform for continuous parameters. Options are
                - "none" (default): keep values unchanged.
                - "normalize": transforms the domain bounds to [0, 1]
            discrete (str, optional): Transform for discrete parameters. Options are
                - "none" (default): keep values unchanged.
                - "normalize": transforms the domain bounds to [0, 1]
            categorical (str, optional): Transform for categoricals. Options are
                - "onehot-encode" (default): A parameter with levels [A, B, C] transforms to 3 columns holding values [0, 1]
                - "dummy-encode": a parameter with levels [A, B, C] transforms to 2 columns holding values [0, 1]
                - "label-encode": a parameter with levels [A, B, C] transfroms to 1 columns with values [0, 1, 2]
                - "none": keep values unchanged

        Raises:
            ValueError: Unknown transform.

        Returns:
            pd.DataFrame: Transformed points. Columns that don't correspond to parameters are dropped.
        """
        transformed = []
        for p in self:
            s = points[p.name]
            if isinstance(p, Continuous):
                if continuous == "none":
                    transformed.append(s)
                elif continuous == "normalize":
                    transformed.append(p.to_unit_range(s))
                else:
                    raise ValueError(f"Unknown continuous transform {continuous}.")
            if isinstance(p, Discrete):
                if discrete == "none":
                    transformed.append(s)
                elif discrete == "normalize":
                    transformed.append(p.to_unit_range(s))
                else:
                    raise ValueError(f"Unknown discrete transform {discrete}.")
            if isinstance(p, Categorical):
                if categorical == "none":
                    transformed.append(s)
                elif categorical == "onehot-encode":
                    transformed.append(p.to_onehot_encoding(s))
                elif categorical == "dummy-encode":
                    transformed.append(p.to_dummy_encoding(s))
                elif categorical == "label-encode":
                    transformed.append(p.to_label_encoding(s))
                else:
                    raise ValueError(f"Unknown categorical transform {categorical}.")
        return pd.concat(transformed, axis=1)

    def to_config(self) -> List[dict]:
        """Configuration of the parameter space."""
        return [param.to_config() for param in self.parameters.values()]

    def get(self, types) -> "Parameters":
        """Get all parameters of the given type(s)."""
        return Parameters([p for p in self if isinstance(p, types)])

    def to_df(self, x: np.ndarray, to_numeric=False) -> pd.DataFrame:
        """Create a dataframe for a given numpy array of parameter values."""
        X = pd.DataFrame(np.atleast_2d(x), columns=self.names)
        if to_numeric:
            for n in self.get((Continuous, Discrete)).names:
                X[n] = pd.to_numeric(X[n])
        return X

bounds property

Return the parameter bounds.

__init__(parameters)

It can be constructed either from a list / tuple (of at least one) Parameter objects

Parameters([
    Continuous(name="foo", domain=[1, 10]),
    Discrete(name="bar", domain=[1, 2, 3, 4]),
    Categorical(name="baz", domain=["A", "B", 3]),
])
or from a list / tuple of dicts
Parameters([
    {"name": "foo", "type": "continuous", "domain": [1, 10]},
    {"name": "bar", "type": "discrete", "domain": [1, 2, 3, 4]},
    {"name": "baz", "type": "categorical", "domain": ["A", "B", 3]},
    {"name": "baz", "type": "categorical", "domain": ["A", "B", 3], extra="info"},
])
In particular, Parameters().init and the to_config method of each Parameter type are inverses. Parameters(conf).to_config() == conf

Source code in opti/parameter.py
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
def __init__(self, parameters):
    """
    It can be constructed either from a list / tuple (of at least one) Parameter objects
    ```
    Parameters([
        Continuous(name="foo", domain=[1, 10]),
        Discrete(name="bar", domain=[1, 2, 3, 4]),
        Categorical(name="baz", domain=["A", "B", 3]),
    ])
    ```
    or from a list / tuple of dicts
    ```
    Parameters([
        {"name": "foo", "type": "continuous", "domain": [1, 10]},
        {"name": "bar", "type": "discrete", "domain": [1, 2, 3, 4]},
        {"name": "baz", "type": "categorical", "domain": ["A", "B", 3]},
        {"name": "baz", "type": "categorical", "domain": ["A", "B", 3], extra="info"},
    ])
    ```
    In particular, Parameters().__init__ and the to_config method of each Parameter type are inverses.
    Parameters(conf).to_config() == conf
    """
    if not isinstance(parameters, (list, tuple)):
        raise TypeError("Parameters expects a list or tuple of parameters.")

    self.parameters = {}
    for d in parameters:
        if not isinstance(d, Parameter):
            d = make_parameter(**d)
        if d.name in self.parameters:
            raise ValueError(f"Duplicate parameter name {d.name}")
        self.parameters[d.name] = d

contains(points)

Check if points are inside the domain of each parameter.

Source code in opti/parameter.py
418
419
420
421
422
423
def contains(self, points: pd.DataFrame) -> pd.Series:
    """Check if points are inside the domain of each parameter."""
    if isinstance(points, pd.DataFrame):
        points = points[self.names]
    b = np.stack([self[k].contains(v) for k, v in points.items()], axis=1)
    return b.all(axis=1)

get(types)

Get all parameters of the given type(s).

Source code in opti/parameter.py
496
497
498
def get(self, types) -> "Parameters":
    """Get all parameters of the given type(s)."""
    return Parameters([p for p in self if isinstance(p, types)])

round(points)

Round points to the closest contained values.

Source code in opti/parameter.py
425
426
427
def round(self, points: pd.DataFrame) -> pd.DataFrame:
    """Round points to the closest contained values."""
    return pd.concat([self[k].round(v) for k, v in points.items()], axis=1)

sample(n=1)

Draw uniformly distributed random samples.

Source code in opti/parameter.py
429
430
431
def sample(self, n: int = 1) -> pd.DataFrame:
    """Draw uniformly distributed random samples."""
    return pd.concat([param.sample(n) for param in self], axis=1)

to_config()

Configuration of the parameter space.

Source code in opti/parameter.py
492
493
494
def to_config(self) -> List[dict]:
    """Configuration of the parameter space."""
    return [param.to_config() for param in self.parameters.values()]

to_df(x, to_numeric=False)

Create a dataframe for a given numpy array of parameter values.

Source code in opti/parameter.py
500
501
502
503
504
505
506
def to_df(self, x: np.ndarray, to_numeric=False) -> pd.DataFrame:
    """Create a dataframe for a given numpy array of parameter values."""
    X = pd.DataFrame(np.atleast_2d(x), columns=self.names)
    if to_numeric:
        for n in self.get((Continuous, Discrete)).names:
            X[n] = pd.to_numeric(X[n])
    return X

transform(points, continuous='none', discrete='none', categorical='onehot-encode')

Transfrom the given dataframe according to a set of transformation rules.

Parameters:

Name Type Description Default
points DataFrame

Dataframe to transfrom. Must contain columns for each parameter and may contain additional columns.

required
continuous str

Transform for continuous parameters. Options are - "none" (default): keep values unchanged. - "normalize": transforms the domain bounds to [0, 1]

'none'
discrete str

Transform for discrete parameters. Options are - "none" (default): keep values unchanged. - "normalize": transforms the domain bounds to [0, 1]

'none'
categorical str

Transform for categoricals. Options are - "onehot-encode" (default): A parameter with levels [A, B, C] transforms to 3 columns holding values [0, 1] - "dummy-encode": a parameter with levels [A, B, C] transforms to 2 columns holding values [0, 1] - "label-encode": a parameter with levels [A, B, C] transfroms to 1 columns with values [0, 1, 2] - "none": keep values unchanged

'onehot-encode'

Raises:

Type Description
ValueError

Unknown transform.

Returns:

Type Description
DataFrame

pd.DataFrame: Transformed points. Columns that don't correspond to parameters are dropped.

Source code in opti/parameter.py
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
def transform(
    self,
    points: pd.DataFrame,
    continuous: str = "none",
    discrete: str = "none",
    categorical: str = "onehot-encode",
) -> pd.DataFrame:
    """Transfrom the given dataframe according to a set of transformation rules.

    Args:
        points (pd.DataFrame): Dataframe to transfrom. Must contain columns for each parameter and may contain additional columns.
        continuous (str, optional): Transform for continuous parameters. Options are
            - "none" (default): keep values unchanged.
            - "normalize": transforms the domain bounds to [0, 1]
        discrete (str, optional): Transform for discrete parameters. Options are
            - "none" (default): keep values unchanged.
            - "normalize": transforms the domain bounds to [0, 1]
        categorical (str, optional): Transform for categoricals. Options are
            - "onehot-encode" (default): A parameter with levels [A, B, C] transforms to 3 columns holding values [0, 1]
            - "dummy-encode": a parameter with levels [A, B, C] transforms to 2 columns holding values [0, 1]
            - "label-encode": a parameter with levels [A, B, C] transfroms to 1 columns with values [0, 1, 2]
            - "none": keep values unchanged

    Raises:
        ValueError: Unknown transform.

    Returns:
        pd.DataFrame: Transformed points. Columns that don't correspond to parameters are dropped.
    """
    transformed = []
    for p in self:
        s = points[p.name]
        if isinstance(p, Continuous):
            if continuous == "none":
                transformed.append(s)
            elif continuous == "normalize":
                transformed.append(p.to_unit_range(s))
            else:
                raise ValueError(f"Unknown continuous transform {continuous}.")
        if isinstance(p, Discrete):
            if discrete == "none":
                transformed.append(s)
            elif discrete == "normalize":
                transformed.append(p.to_unit_range(s))
            else:
                raise ValueError(f"Unknown discrete transform {discrete}.")
        if isinstance(p, Categorical):
            if categorical == "none":
                transformed.append(s)
            elif categorical == "onehot-encode":
                transformed.append(p.to_onehot_encoding(s))
            elif categorical == "dummy-encode":
                transformed.append(p.to_dummy_encoding(s))
            elif categorical == "label-encode":
                transformed.append(p.to_label_encoding(s))
            else:
                raise ValueError(f"Unknown categorical transform {categorical}.")
    return pd.concat(transformed, axis=1)

make_parameter(name, type, domain=None, **kwargs)

Make a parameter object from a configuration

p = make_parameter(**config)

Parameters:

Name Type Description Default
type str

"continuous", "discrete" or "categorical"

required
name str

Name of the parameter

required
domain list

Domain, e.g, [0, 1], [1, 2.5, 5] or ["A", "B", "C"]

None
Source code in opti/parameter.py
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
def make_parameter(
    name: str,
    type: str,
    domain: Optional[Sequence] = None,
    **kwargs,
):
    """Make a parameter object from a configuration

    p = make_parameter(**config)

    Args:
        type (str): "continuous", "discrete" or "categorical"
        name (str): Name of the parameter
        domain (list): Domain, e.g, [0, 1], [1, 2.5, 5] or ["A", "B", "C"]
    """
    parameter = {
        "continuous": Continuous,
        "discrete": Discrete,
        "categorical": Categorical,
    }[type.lower()]
    if domain is None and parameter is not Continuous:
        raise ValueError(f"Domain not given for parameter {name}.")

    return parameter(name=name, domain=domain, **kwargs)