Skip to content

Parameter

Collection of possible parameter specifications for the distribution objects.

Example choices defined:

Identity: f = x LinearCombination: f = X @ beta + Y @ gamma LinearCombinationWithTransform: f = X @ exp(beta) + Y @ gamma ScaledMatrix f = lam * P MixtureParameterVector f= X[I] MixtureParameterMatrix f= np.diag(lam[I])

Parameter dataclass

Bases: ABC

Abstract base class for parameter.

Source code in src/openmcmc/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
@dataclass
class Parameter(ABC):
    """Abstract base class for parameter."""

    @abstractmethod
    def predictor(self, state: dict) -> np.ndarray:
        """Create predictor from the state dictionary using the functional form defined in the specific subclass.

        Args:
            state (dict): dictionary object containing the current state information

        Returns:
            (np.ndarray): predictor vector

        """

    @abstractmethod
    def get_param_list(self) -> list:
        """Extract list of components from parameter specification.

        Returns:
            (list): parameter included as part of predictor

        """

    @abstractmethod
    def get_grad_param_list(self) -> list:
        """Extract list of components from parameter specification that grad is defined for.

        Returns:
            (list): parameter that grad is defined for.

        """

    @abstractmethod
    def grad(self, state: dict, param: str) -> np.ndarray:
        """Compute gradient of single parameter.

        Args:
            state (dict): Dictionary object containing the current state information
            param (str): Compute derivatives WRT this variable

        Returns:
            (np.ndarray): [n_param x n_data] array, gradient with respect to param

        """

predictor(state) abstractmethod

Create predictor from the state dictionary using the functional form defined in the specific subclass.

Parameters:

Name Type Description Default
state dict

dictionary object containing the current state information

 required

Returns:

Type Description
ndarray

predictor vector
Source code in src/openmcmc/parameter.py 
30
31
32
33
34
35
36
37
38
39
40
@abstractmethod
def predictor(self, state: dict) -> np.ndarray:
    """Create predictor from the state dictionary using the functional form defined in the specific subclass.

    Args:
        state (dict): dictionary object containing the current state information

    Returns:
        (np.ndarray): predictor vector

    """

get_param_list() abstractmethod

Extract list of components from parameter specification.

Returns:

Type Description
list

parameter included as part of predictor
Source code in src/openmcmc/parameter.py 
42
43
44
45
46
47
48
49
@abstractmethod
def get_param_list(self) -> list:
    """Extract list of components from parameter specification.

    Returns:
        (list): parameter included as part of predictor

    """

get_grad_param_list() abstractmethod

Extract list of components from parameter specification that grad is defined for.

Returns:

Type Description
list

parameter that grad is defined for.
Source code in src/openmcmc/parameter.py 
51
52
53
54
55
56
57
58
@abstractmethod
def get_grad_param_list(self) -> list:
    """Extract list of components from parameter specification that grad is defined for.

    Returns:
        (list): parameter that grad is defined for.

    """

grad(state, param) abstractmethod

Compute gradient of single parameter.

Parameters:

Name Type Description Default
state dict

Dictionary object containing the current state information

 required
param str

Compute derivatives WRT this variable

 required

Returns:

Type Description
ndarray

[n_param x n_data] array, gradient with respect to param
Source code in src/openmcmc/parameter.py 
60
61
62
63
64
65
66
67
68
69
70
71
@abstractmethod
def grad(self, state: dict, param: str) -> np.ndarray:
    """Compute gradient of single parameter.

    Args:
        state (dict): Dictionary object containing the current state information
        param (str): Compute derivatives WRT this variable

    Returns:
        (np.ndarray): [n_param x n_data] array, gradient with respect to param

    """

Identity dataclass

Bases: Parameter

Class specifying a simple predictor in a single term.

Predictor has the functional form

f = x

The gradient should only be used for scalar and vector inputs

Parameters:

Name Type Description Default
form str

string specifying the element of state which determines the parameter

 required

Attributes:

Name Type Description
form str

string specifying the element of state which determines the parameter.
Source code in src/openmcmc/parameter.py 
 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
129
130
131
132
133
134
135
136
137
138
139
140
141
@dataclass
class Identity(Parameter):
    """Class specifying a simple predictor in a single term.

    Predictor has the functional form:
        f = x

    The gradient should only be used for scalar and vector inputs

    Args:
        form (str): string specifying the element of state which determines the parameter

    Attributes:
        form (str): string specifying the element of state which determines the parameter.

    """

    form: str

    def predictor(self, state: dict) -> np.ndarray:
        """Create predictor from the state dictionary using the functional form defined in the specific subclass.

        Args:
            state (dict): dictionary object containing the current state information

        Returns:
            (np.ndarray): predictor vector

        """
        return state[self.form]

    def get_param_list(self) -> list:
        """Extract list of components from parameter specification that grad is defined for.

        Returns:
            (list): parameter that grad is defined for.

        """
        return [self.form]

    def get_grad_param_list(self) -> list:
        """Extract list of components from parameter specification that grad is defined for.

        Returns:
            (list): parameter that grad is defined for.

        """
        return [self.form]

    def grad(self, state: dict, param: str) -> np.ndarray:
        """Compute gradient of single parameter.

        Args:
            state (dict): Dictionary object containing the current state information
            param (str): Compute derivatives WRT this variable

        Returns:
            (np.ndarray): [n_param x n_data] array, gradient with respect to param

        """
        if state[self.form].shape[1] > 1:
            raise ValueError("Gradient in Identity should not be used for variables 2D and above.")
        p = state[self.form].size
        if param == self.form:
            grad = np.eye(p)
        else:
            grad = np.zeros(shape=(p, p))
        return grad

predictor(state)

Create predictor from the state dictionary using the functional form defined in the specific subclass.

Parameters:

Name Type Description Default
state dict

dictionary object containing the current state information

 required

Returns:

Type Description
ndarray

predictor vector
Source code in src/openmcmc/parameter.py 
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
def predictor(self, state: dict) -> np.ndarray:
    """Create predictor from the state dictionary using the functional form defined in the specific subclass.

    Args:
        state (dict): dictionary object containing the current state information

    Returns:
        (np.ndarray): predictor vector

    """
    return state[self.form]

get_param_list()

Extract list of components from parameter specification that grad is defined for.

Returns:

Type Description
list

parameter that grad is defined for.
Source code in src/openmcmc/parameter.py 
105
106
107
108
109
110
111
112
def get_param_list(self) -> list:
    """Extract list of components from parameter specification that grad is defined for.

    Returns:
        (list): parameter that grad is defined for.

    """
    return [self.form]

get_grad_param_list()

Extract list of components from parameter specification that grad is defined for.

Returns:

Type Description
list

parameter that grad is defined for.
Source code in src/openmcmc/parameter.py 
114
115
116
117
118
119
120
121
def get_grad_param_list(self) -> list:
    """Extract list of components from parameter specification that grad is defined for.

    Returns:
        (list): parameter that grad is defined for.

    """
    return [self.form]

grad(state, param)

Compute gradient of single parameter.

Parameters:

Name Type Description Default
state dict

Dictionary object containing the current state information

 required
param str

Compute derivatives WRT this variable

 required

Returns:

Type Description
ndarray

[n_param x n_data] array, gradient with respect to param
Source code in src/openmcmc/parameter.py 
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
def grad(self, state: dict, param: str) -> np.ndarray:
    """Compute gradient of single parameter.

    Args:
        state (dict): Dictionary object containing the current state information
        param (str): Compute derivatives WRT this variable

    Returns:
        (np.ndarray): [n_param x n_data] array, gradient with respect to param

    """
    if state[self.form].shape[1] > 1:
        raise ValueError("Gradient in Identity should not be used for variables 2D and above.")
    p = state[self.form].size
    if param == self.form:
        grad = np.eye(p)
    else:
        grad = np.zeros(shape=(p, p))
    return grad

LinearCombination dataclass

Bases: Parameter

Class specifying linear combination form .

This Parameter type is typically in the mean of a Normal distribution in a linear regression type case.

Predictor has the form predictor = sum_i (value[i] @ key[i]) using the form dictionary input

Attributes:

Name Type Description
form dict

dict specifying the term and prefactor in the linear combination. example: {'beta': 'X', 'alpha': 'A'} produces linear combination X @ beta + A @ alpha.
Source code in src/openmcmc/parameter.py 
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
221
222
223
224
225
226
227
228
@dataclass
class LinearCombination(Parameter):
    """Class specifying linear combination form .

    This Parameter type is typically in the mean of a Normal distribution in a linear regression type case.

    Predictor has the form
        predictor  = sum_i (value[i] @ key[i])
    using the form dictionary input

    Attributes:
        form (dict): dict specifying the term and prefactor in the linear combination.
            example: {'beta': 'X', 'alpha': 'A'} produces linear combination X @ beta + A @ alpha.

    """

    form: dict

    def predictor(self, state: dict) -> np.ndarray:
        """Create predictor from the state dictionary using the functional form defined in the specific subclass.

        Args:
            state (dict): dictionary object containing the current state information

        Returns:
            (np.ndarray): predictor vector

        """
        return self.predictor_conditional(state)

    def predictor_conditional(self, state: dict, term_to_exclude: Union[str, list] = None) -> np.ndarray:
        """Extract predictor from the state dictionary using the functional form defined in the specific subclass excluding parameters.

        Used when estimating conditional distributions of those parameters.

        Args:
            state (dict): dictionary object containing the current state information
            term_to_exclude (Union[str, list]): terms to exclude from predictor

        Returns:
            (np.ndarray): predictor vector

        """
        if term_to_exclude is None:
            term_to_exclude = []

        if isinstance(term_to_exclude, str):
            term_to_exclude = [term_to_exclude]

        sum_terms = 0
        for prm, prefactor in self.form.items():
            if prm not in term_to_exclude:
                sum_terms += state[prefactor] @ state[prm]
        return sum_terms

    def get_param_list(self) -> list:
        """Extract list of components from parameter specification that grad is defined for.

        Returns:
            (list): parameter that grad is defined for.

        """
        return list(self.form.keys()) + list(self.form.values())

    def get_grad_param_list(self) -> list:
        """Extract list of components from parameter specification that grad is defined for.

        Returns:
            (list): parameter that grad is defined for.

        """
        return list(self.form.keys())

    def grad(self, state: dict, param: str) -> np.ndarray:
        """Compute gradient of single parameter.

        Args:
            state (dict): Dictionary object containing the current state information
            param (str): Compute derivatives WRT this variable

        Returns:
            (np.ndarray): [n_param x n_data] array, gradient with respect to param

        """
        return state[self.form[param]].T

predictor(state)

Create predictor from the state dictionary using the functional form defined in the specific subclass.

Parameters:

Name Type Description Default
state dict

dictionary object containing the current state information

 required

Returns:

Type Description
ndarray

predictor vector
Source code in src/openmcmc/parameter.py 
162
163
164
165
166
167
168
169
170
171
172
def predictor(self, state: dict) -> np.ndarray:
    """Create predictor from the state dictionary using the functional form defined in the specific subclass.

    Args:
        state (dict): dictionary object containing the current state information

    Returns:
        (np.ndarray): predictor vector

    """
    return self.predictor_conditional(state)

predictor_conditional(state, term_to_exclude=None)

Extract predictor from the state dictionary using the functional form defined in the specific subclass excluding parameters.

Used when estimating conditional distributions of those parameters.

Parameters:

Name Type Description Default
state dict

dictionary object containing the current state information

 required
term_to_exclude Union[str, list]

terms to exclude from predictor

 None

Returns:

Type Description
ndarray

predictor vector
Source code in src/openmcmc/parameter.py 
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
def predictor_conditional(self, state: dict, term_to_exclude: Union[str, list] = None) -> np.ndarray:
    """Extract predictor from the state dictionary using the functional form defined in the specific subclass excluding parameters.

    Used when estimating conditional distributions of those parameters.

    Args:
        state (dict): dictionary object containing the current state information
        term_to_exclude (Union[str, list]): terms to exclude from predictor

    Returns:
        (np.ndarray): predictor vector

    """
    if term_to_exclude is None:
        term_to_exclude = []

    if isinstance(term_to_exclude, str):
        term_to_exclude = [term_to_exclude]

    sum_terms = 0
    for prm, prefactor in self.form.items():
        if prm not in term_to_exclude:
            sum_terms += state[prefactor] @ state[prm]
    return sum_terms

get_param_list()

Extract list of components from parameter specification that grad is defined for.

Returns:

Type Description
list

parameter that grad is defined for.
Source code in src/openmcmc/parameter.py 
199
200
201
202
203
204
205
206
def get_param_list(self) -> list:
    """Extract list of components from parameter specification that grad is defined for.

    Returns:
        (list): parameter that grad is defined for.

    """
    return list(self.form.keys()) + list(self.form.values())

get_grad_param_list()

Extract list of components from parameter specification that grad is defined for.

Returns:

Type Description
list

parameter that grad is defined for.
Source code in src/openmcmc/parameter.py 
208
209
210
211
212
213
214
215
def get_grad_param_list(self) -> list:
    """Extract list of components from parameter specification that grad is defined for.

    Returns:
        (list): parameter that grad is defined for.

    """
    return list(self.form.keys())

grad(state, param)

Compute gradient of single parameter.

Parameters:

Name Type Description Default
state dict

Dictionary object containing the current state information

 required
param str

Compute derivatives WRT this variable

 required

Returns:

Type Description
ndarray

[n_param x n_data] array, gradient with respect to param
Source code in src/openmcmc/parameter.py 
217
218
219
220
221
222
223
224
225
226
227
228
def grad(self, state: dict, param: str) -> np.ndarray:
    """Compute gradient of single parameter.

    Args:
        state (dict): Dictionary object containing the current state information
        param (str): Compute derivatives WRT this variable

    Returns:
        (np.ndarray): [n_param x n_data] array, gradient with respect to param

    """
    return state[self.form[param]].T

LinearCombinationWithTransform dataclass

Bases: LinearCombination

Linear combination of parameters from the state, with optional exponential transformation for the parameter elements.

Currently, the only allowed transformation is the exponential transform.

This Parameter type is typically in the mean of a Normal distribution and could be used to impose positivity of the parameters

Predictor has the form predictor = sum_i (value[i] @ transform(key[i])) using the form dictionary input

Attributes:

Name Type Description
transform dict

dict with logicals specifying whether exp(.) transform should be applied to parameter example: form={'beta': X}, transform={'beta': True} will produce X @ np.exp(beta)
Source code in src/openmcmc/parameter.py 
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
@dataclass
class LinearCombinationWithTransform(LinearCombination):
    """Linear combination of parameters from the state, with optional exponential transformation for the parameter elements.

    Currently, the only allowed transformation is the exponential transform.

    This Parameter type is typically in the mean of a Normal distribution and could be
    used to impose positivity of the parameters

    Predictor has the form
        predictor  = sum_i (value[i] @ transform(key[i]))
    using the form dictionary input

    Attributes:
        transform (dict): dict with logicals specifying whether exp(.) transform should
            be applied to parameter
            example: form={'beta': X}, transform={'beta': True} will produce X @ np.exp(beta)

    """

    transform: dict

    def predictor_conditional(self, state: dict, term_to_exclude: Union[str, list] = None) -> np.ndarray:
        """Extract predictor from the state dictionary using the functional form defined in the specific subclass excluding parameters.

        Used when estimating conditional distributions of those parameters.

        Args:
            state (dict): dictionary object containing the current state information
            term_to_exclude (list): terms to exclude from predictor

        Returns:
            (np.ndarray): predictor vector

        """
        if term_to_exclude is None:
            term_to_exclude = []

        if isinstance(term_to_exclude, str):
            term_to_exclude = [term_to_exclude]

        sum_terms = 0
        for prm, prefactor in self.form.items():
            if prm not in term_to_exclude:
                param = state[prm]
                if self.transform[prm]:
                    param = np.exp(param)
                sum_terms += state[prefactor] @ param
        return sum_terms

    def grad(self, state: dict, param: str) -> np.ndarray:
        """Compute gradient of single parameter.

        Args:
            state (dict): Dictionary object containing the current state information
            param (str): Compute derivatives WRT this variable

        Returns:
            (np.ndarray): [n_param x n_data] array, gradient with respect to param

        """
        if self.transform[param]:
            if sparse.issparse(state[self.form[param]]):
                return state[self.form[param]].multiply(np.exp(state[param]).flatten()).T
            return np.exp(state[param]) * (state[self.form[param]].T)

        return state[self.form[param]].T

predictor_conditional(state, term_to_exclude=None)

Extract predictor from the state dictionary using the functional form defined in the specific subclass excluding parameters.

Used when estimating conditional distributions of those parameters.

Parameters:

Name Type Description Default
state dict

dictionary object containing the current state information

 required
term_to_exclude list

terms to exclude from predictor

 None

Returns:

Type Description
ndarray

predictor vector
Source code in src/openmcmc/parameter.py 
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
def predictor_conditional(self, state: dict, term_to_exclude: Union[str, list] = None) -> np.ndarray:
    """Extract predictor from the state dictionary using the functional form defined in the specific subclass excluding parameters.

    Used when estimating conditional distributions of those parameters.

    Args:
        state (dict): dictionary object containing the current state information
        term_to_exclude (list): terms to exclude from predictor

    Returns:
        (np.ndarray): predictor vector

    """
    if term_to_exclude is None:
        term_to_exclude = []

    if isinstance(term_to_exclude, str):
        term_to_exclude = [term_to_exclude]

    sum_terms = 0
    for prm, prefactor in self.form.items():
        if prm not in term_to_exclude:
            param = state[prm]
            if self.transform[prm]:
                param = np.exp(param)
            sum_terms += state[prefactor] @ param
    return sum_terms

grad(state, param)

Compute gradient of single parameter.

Parameters:

Name Type Description Default
state dict

Dictionary object containing the current state information

 required
param str

Compute derivatives WRT this variable

 required

Returns:

Type Description
ndarray

[n_param x n_data] array, gradient with respect to param
Source code in src/openmcmc/parameter.py 
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
def grad(self, state: dict, param: str) -> np.ndarray:
    """Compute gradient of single parameter.

    Args:
        state (dict): Dictionary object containing the current state information
        param (str): Compute derivatives WRT this variable

    Returns:
        (np.ndarray): [n_param x n_data] array, gradient with respect to param

    """
    if self.transform[param]:
        if sparse.issparse(state[self.form[param]]):
            return state[self.form[param]].multiply(np.exp(state[param]).flatten()).T
        return np.exp(state[param]) * (state[self.form[param]].T)

    return state[self.form[param]].T

ScaledMatrix dataclass

Bases: Parameter

Defines parameter a scalar factor in front of a matrix.

This is often used in case where we have a scalar variance in front of an unscaled precision matrix. Where we have a gamma distribution for the scalar parameter which wish to estimate

Linear combinations have the form

predictor = scalar * matrix

Attributes:

Name Type Description
matrix str

variable name of the un-scaled matrix
scalar str

variable name of the scalar term
Source code in src/openmcmc/parameter.py 
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
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
@dataclass
class ScaledMatrix(Parameter):
    """Defines parameter a scalar factor in front of a matrix.

    This is often used in case where we have a scalar variance in front of an unscaled precision matrix.
    Where we have a gamma distribution for the scalar parameter which wish to estimate

    Linear combinations have the form:
        predictor = scalar * matrix

    Attributes:
        matrix (str): variable name of the un-scaled matrix
        scalar (str): variable name of the scalar term

    """

    matrix: str
    scalar: str

    def predictor(self, state: dict) -> np.ndarray:
        """Create predictor from the state dictionary using the functional form defined in the specific subclass.

        Args:
            state (dict): dictionary object containing the current state information

        Returns:
            (np.ndarray): predictor vector

        """
        return float(state[self.scalar].item()) * state[self.matrix]

    def get_param_list(self) -> list:
        """Extract list of components from parameter specification that grad is defined for.

        Returns:
            (list): parameter that grad is defined for.

        """
        return [self.scalar, self.matrix]

    def get_grad_param_list(self) -> list:
        """Extract list of components from parameter specification that grad is defined for.

        Returns:
            (list): parameter that grad is defined for.

        """
        return [self.scalar]

    def grad(self, state: dict, param: str) -> np.ndarray:
        """Compute gradient of single parameter.

        Args:
            state (dict): Dictionary object containing the current state information
            param (str): Compute derivatives WRT this variable

        Returns:
            (np.ndarray): [n_param x n_data] array, gradient with respect to param

        """
        return state[self.matrix]

    def precision_unscaled(self, state: dict, _) -> np.ndarray:
        """Return the precision matrix un-scaled by the scalar precision parameter.

        Args:
            state (dict): state dictionary
            _ (int): argument unused but matches with version in MixtureParameterMatrix where element is needed

        Returns:
            (np.ndarray): unscaled precision matrix

        """
        return state[self.matrix]

predictor(state)

Create predictor from the state dictionary using the functional form defined in the specific subclass.

Parameters:

Name Type Description Default
state dict

dictionary object containing the current state information

 required

Returns:

Type Description
ndarray

predictor vector
Source code in src/openmcmc/parameter.py 
319
320
321
322
323
324
325
326
327
328
329
def predictor(self, state: dict) -> np.ndarray:
    """Create predictor from the state dictionary using the functional form defined in the specific subclass.

    Args:
        state (dict): dictionary object containing the current state information

    Returns:
        (np.ndarray): predictor vector

    """
    return float(state[self.scalar].item()) * state[self.matrix]

get_param_list()

Extract list of components from parameter specification that grad is defined for.

Returns:

Type Description
list

parameter that grad is defined for.
Source code in src/openmcmc/parameter.py 
331
332
333
334
335
336
337
338
def get_param_list(self) -> list:
    """Extract list of components from parameter specification that grad is defined for.

    Returns:
        (list): parameter that grad is defined for.

    """
    return [self.scalar, self.matrix]

get_grad_param_list()

Extract list of components from parameter specification that grad is defined for.

Returns:

Type Description
list

parameter that grad is defined for.
Source code in src/openmcmc/parameter.py 
340
341
342
343
344
345
346
347
def get_grad_param_list(self) -> list:
    """Extract list of components from parameter specification that grad is defined for.

    Returns:
        (list): parameter that grad is defined for.

    """
    return [self.scalar]

grad(state, param)

Compute gradient of single parameter.

Parameters:

Name Type Description Default
state dict

Dictionary object containing the current state information

 required
param str

Compute derivatives WRT this variable

 required

Returns:

Type Description
ndarray

[n_param x n_data] array, gradient with respect to param
Source code in src/openmcmc/parameter.py 
349
350
351
352
353
354
355
356
357
358
359
360
def grad(self, state: dict, param: str) -> np.ndarray:
    """Compute gradient of single parameter.

    Args:
        state (dict): Dictionary object containing the current state information
        param (str): Compute derivatives WRT this variable

    Returns:
        (np.ndarray): [n_param x n_data] array, gradient with respect to param

    """
    return state[self.matrix]

precision_unscaled(state, _)

Return the precision matrix un-scaled by the scalar precision parameter.

Parameters:

Name Type Description Default
state dict

state dictionary

 required
_ int

argument unused but matches with version in MixtureParameterMatrix where element is needed

 required

Returns:

Type Description
ndarray

unscaled precision matrix
Source code in src/openmcmc/parameter.py 
362
363
364
365
366
367
368
369
370
371
372
373
def precision_unscaled(self, state: dict, _) -> np.ndarray:
    """Return the precision matrix un-scaled by the scalar precision parameter.

    Args:
        state (dict): state dictionary
        _ (int): argument unused but matches with version in MixtureParameterMatrix where element is needed

    Returns:
        (np.ndarray): unscaled precision matrix

    """
    return state[self.matrix]

MixtureParameter dataclass

Bases: Parameter, ABC

Abstract Parameter class for a mixture distribution.

Subclasses implemented for both:

  • vector-valued parameter (MixtureParameterVector)
  • diagonal matrix-valued parameter (MixtureParameterMatrix) where the elements of the vector or matrix diagonal are allocated based on the allocation parameter.
Source code in src/openmcmc/parameter.py 
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
@dataclass
class MixtureParameter(Parameter, ABC):
    """Abstract Parameter class for a mixture distribution.

    Subclasses implemented for both:

    - vector-valued parameter (MixtureParameterVector)
    - diagonal matrix-valued parameter (MixtureParameterMatrix)
    where the elements of the vector or matrix diagonal are allocated based
    on the allocation parameter.

    """

    param: str
    allocation: str

    def get_element_match(self, state: dict, element_index: Union[int, np.ndarray]) -> np.ndarray:
        """Extract the parts of self.allocation which have given element number.

        used in the gradient function to pull out gradient for given element.

        Args:
            state (dict): state vector
            element_index (int, np.array): element index or set of integers

        Returns:
            (np.array(dtype=int)): element matches with 1 where there is a match and 0 where there isn't

        """
        if isinstance(element_index, np.ndarray) and element_index.size > 1:
            element_index = element_index.reshape((1, -1))

        return np.array(state[self.allocation] == element_index, dtype=int, ndmin=2)

    def get_param_list(self) -> list:
        """Extract list of components from parameter specification that grad is defined for.

        Returns:
            (list): parameter that grad is defined for.

        """
        return [self.param, self.allocation]

get_element_match(state, element_index)

Extract the parts of self.allocation which have given element number.

used in the gradient function to pull out gradient for given element.

Parameters:

Name Type Description Default
state dict

state vector

 required
element_index (int, array)

element index or set of integers

 required

Returns:

Type Description
array(dtype=int)

element matches with 1 where there is a match and 0 where there isn't
Source code in src/openmcmc/parameter.py 
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
def get_element_match(self, state: dict, element_index: Union[int, np.ndarray]) -> np.ndarray:
    """Extract the parts of self.allocation which have given element number.

    used in the gradient function to pull out gradient for given element.

    Args:
        state (dict): state vector
        element_index (int, np.array): element index or set of integers

    Returns:
        (np.array(dtype=int)): element matches with 1 where there is a match and 0 where there isn't

    """
    if isinstance(element_index, np.ndarray) and element_index.size > 1:
        element_index = element_index.reshape((1, -1))

    return np.array(state[self.allocation] == element_index, dtype=int, ndmin=2)

get_param_list()

Extract list of components from parameter specification that grad is defined for.

Returns:

Type Description
list

parameter that grad is defined for.
Source code in src/openmcmc/parameter.py 
410
411
412
413
414
415
416
417
def get_param_list(self) -> list:
    """Extract list of components from parameter specification that grad is defined for.

    Returns:
        (list): parameter that grad is defined for.

    """
    return [self.param, self.allocation]

MixtureParameterVector dataclass

Bases: MixtureParameter

Vector parameter: elements of the vector are obtained from sub-parameter 'param' according to the allocation.

The allocation parameter defines a mapping between a R^m and R^n where typically m<=n and m is the number true underlying number of parameters in the model but due to the representation/algebra in other parts of the model this is expanded out to an n parameter model where the values of m are copied according to the index vector

predictor = param [allocation]

Attributes:

Name Type Description
param str

name of underlying state component used to generate parameter.
allocation ndarray

name of allocation parameter within state dict.
Source code in src/openmcmc/parameter.py 
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
@dataclass
class MixtureParameterVector(MixtureParameter):
    """Vector parameter: elements of the vector are obtained from sub-parameter 'param' according to the allocation.

    The allocation parameter defines a mapping between a R^m and R^n where typically m<=n and m is the
    number true underlying number of parameters in the model but due to the representation/algebra in
    other parts of the model this is expanded out to an n parameter model where the values of m are copied
    according to the index vector

    predictor = param [allocation]

    Attributes:
        param (str): name of underlying state component used to generate parameter.
        allocation (np.ndarray): name of allocation parameter within state dict.

    """

    def predictor(self, state: dict) -> np.ndarray:
        """Create predictor from the state dictionary using the functional form defined in the specific subclass.

        Args:
            state (dict): dictionary object containing the current state information

        Returns:
            (np.ndarray): predictor vector

        """
        return state[self.param][state[self.allocation].flatten()]

    def grad(self, state: dict, param: str):
        """Compute gradient of single parameter.

        Args:
            state (dict): Dictionary object containing the current state information
            param (str): Compute derivatives WRT this variable

        Returns:
            (np.ndarray): [n_param x n_data] array, gradient with respect to param

        """
        element_index = np.arange(0, state[param].size)

        return self.get_element_match(state, element_index).astype(np.float64).T

    def get_grad_param_list(self) -> list:
        """Extract list of components from parameter specification that grad is defined for.

        Returns:
            (list): parameter that grad is defined for.

        """
        return [self.param]

predictor(state)

Create predictor from the state dictionary using the functional form defined in the specific subclass.

Parameters:

Name Type Description Default
state dict

dictionary object containing the current state information

 required

Returns:

Type Description
ndarray

predictor vector
Source code in src/openmcmc/parameter.py 
437
438
439
440
441
442
443
444
445
446
447
def predictor(self, state: dict) -> np.ndarray:
    """Create predictor from the state dictionary using the functional form defined in the specific subclass.

    Args:
        state (dict): dictionary object containing the current state information

    Returns:
        (np.ndarray): predictor vector

    """
    return state[self.param][state[self.allocation].flatten()]

grad(state, param)

Compute gradient of single parameter.

Parameters:

Name Type Description Default
state dict

Dictionary object containing the current state information

 required
param str

Compute derivatives WRT this variable

 required

Returns:

Type Description
ndarray

[n_param x n_data] array, gradient with respect to param
Source code in src/openmcmc/parameter.py 
449
450
451
452
453
454
455
456
457
458
459
460
461
462
def grad(self, state: dict, param: str):
    """Compute gradient of single parameter.

    Args:
        state (dict): Dictionary object containing the current state information
        param (str): Compute derivatives WRT this variable

    Returns:
        (np.ndarray): [n_param x n_data] array, gradient with respect to param

    """
    element_index = np.arange(0, state[param].size)

    return self.get_element_match(state, element_index).astype(np.float64).T

get_grad_param_list()

Extract list of components from parameter specification that grad is defined for.

Returns:

Type Description
list

parameter that grad is defined for.
Source code in src/openmcmc/parameter.py 
464
465
466
467
468
469
470
471
def get_grad_param_list(self) -> list:
    """Extract list of components from parameter specification that grad is defined for.

    Returns:
        (list): parameter that grad is defined for.

    """
    return [self.param]

MixtureParameterMatrix dataclass

Bases: MixtureParameter

Diagonal matrix parameter: elements of the diagonal are obtained from sub-parameter 'param' according to the allocation index vector.

The allocation parameter defines a mapping between a R^m and R^n where typically m<=n and m is the number true underlying number of parameters in the model but due to the representation/algebra in other parts of the model this is expanded out to an n parameter model where the values of m are copied according to the index vector

predictor = np.diag( param [allocation] )

Attributes:

Name Type Description
param str

name of underlying state component used to generate parameter.
allocation ndarray

name of allocation parameter within state dict.
Source code in src/openmcmc/parameter.py 
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
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
@dataclass
class MixtureParameterMatrix(MixtureParameter):
    """Diagonal matrix parameter: elements of the diagonal are obtained from sub-parameter 'param' according to the allocation index vector.

    The allocation parameter defines a mapping between a R^m and R^n where typically m<=n and m is the
    number true underlying number of parameters in the model but due to the representation/algebra in
    other parts of the model this is expanded out to an n parameter model where the values of m are copied
    according to the index vector

    predictor = np.diag( param [allocation] )

    Attributes:
        param (str): name of underlying state component used to generate parameter.
        allocation (np.ndarray): name of allocation parameter within state dict.

    """

    def predictor(self, state: dict) -> sparse.csc_matrix:
        """Create predictor from the state dictionary using the functional form defined in the specific subclass.

        Args:
            state (dict): dictionary object containing the current state information

        Returns:
            (sparse.csc_matrix): predictor vector

        """
        return sparse.diags(diagonals=state[self.param][state[self.allocation]].flatten(), offsets=0, format="csc")

    def grad(self, state: dict, param: str):
        """Compute gradient of single parameter.

        Args:
            state (dict): Dictionary object containing the current state information
            param (str): Compute derivatives WRT this variable

        Returns:
            (np.ndarray): [n_param x n_data] array, gradient with respect to param

        """
        raise TypeError("Not defined in this case")

    def get_grad_param_list(self) -> list:
        """Extract list of components from parameter specification that grad is defined for.

        Returns:
            (list): parameter that grad is defined for.

        """
        return []

    def precision_unscaled(self, state: dict, element_index: int) -> np.ndarray:
        """Return the precision matrix un-scaled by the scalar precision parameter.

        Args:
            state (dict): state dictionary
            element_index (int): index of element to subset

        Returns:
            (np.ndarray): unscaled precision matrix

        """
        return sparse.diags(diagonals=self.get_element_match(state, element_index).flatten(), offsets=0, format="csc")

predictor(state)

Create predictor from the state dictionary using the functional form defined in the specific subclass.

Parameters:

Name Type Description Default
state dict

dictionary object containing the current state information

 required

Returns:

Type Description
csc_matrix

predictor vector
Source code in src/openmcmc/parameter.py 
491
492
493
494
495
496
497
498
499
500
501
def predictor(self, state: dict) -> sparse.csc_matrix:
    """Create predictor from the state dictionary using the functional form defined in the specific subclass.

    Args:
        state (dict): dictionary object containing the current state information

    Returns:
        (sparse.csc_matrix): predictor vector

    """
    return sparse.diags(diagonals=state[self.param][state[self.allocation]].flatten(), offsets=0, format="csc")

grad(state, param)

Compute gradient of single parameter.

Parameters:

Name Type Description Default
state dict

Dictionary object containing the current state information

 required
param str

Compute derivatives WRT this variable

 required

Returns:

Type Description
ndarray

[n_param x n_data] array, gradient with respect to param
Source code in src/openmcmc/parameter.py 
503
504
505
506
507
508
509
510
511
512
513
514
def grad(self, state: dict, param: str):
    """Compute gradient of single parameter.

    Args:
        state (dict): Dictionary object containing the current state information
        param (str): Compute derivatives WRT this variable

    Returns:
        (np.ndarray): [n_param x n_data] array, gradient with respect to param

    """
    raise TypeError("Not defined in this case")

get_grad_param_list()

Extract list of components from parameter specification that grad is defined for.

Returns:

Type Description
list

parameter that grad is defined for.
Source code in src/openmcmc/parameter.py 
516
517
518
519
520
521
522
523
def get_grad_param_list(self) -> list:
    """Extract list of components from parameter specification that grad is defined for.

    Returns:
        (list): parameter that grad is defined for.

    """
    return []

precision_unscaled(state, element_index)

Return the precision matrix un-scaled by the scalar precision parameter.

Parameters:

Name Type Description Default
state dict

state dictionary

 required
element_index int

index of element to subset

 required

Returns:

Type Description
ndarray

unscaled precision matrix
Source code in src/openmcmc/parameter.py 
525
526
527
528
529
530
531
532
533
534
535
536
def precision_unscaled(self, state: dict, element_index: int) -> np.ndarray:
    """Return the precision matrix un-scaled by the scalar precision parameter.

    Args:
        state (dict): state dictionary
        element_index (int): index of element to subset

    Returns:
        (np.ndarray): unscaled precision matrix

    """
    return sparse.diags(diagonals=self.get_element_match(state, element_index).flatten(), offsets=0, format="csc")