Distributions of Interest¶
The distribution of interest lets us specify the set of samples over which we want our explanations to be faithful. In some cases, we may want to explain the model’s behavior on a particular record, whereas other times we may be interested in a more general behavior over a distribution of samples.
DoI
¶
Bases: AbstractBaseClass
Interface for distributions of interest. The Distribution of Interest (DoI) specifies the samples over which an attribution method is aggregated.
Source code in trulens_explain/trulens/nn/distributions.py
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 |
|
__call__(z, *, model_inputs=None)
abstractmethod
¶
Computes the distribution of interest from an initial point. If z: TensorLike is given, we assume there is only 1 input to the DoI layer. If z: List[TensorLike] is given, it provides all of the inputs to the DoI layer.
Either way, we always return List[List[TensorLike]] (alias Inputs[Uniform[TensorLike]]) with outer list spanning layer inputs, and inner list spanning a distribution's instance.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
z |
OM[Inputs, TensorLike]
|
Input point from which the distribution is derived. If list/tuple, the point is defined by multiple tensors. |
required |
model_inputs |
Optional[ModelInputs]
|
Optional wrapped model input arguments that produce value z at cut. |
None
|
Returns:
Type | Description |
---|---|
OM[Inputs, Uniform[TensorLike]]
|
List of points which are all assigned equal probability mass in the |
OM[Inputs, Uniform[TensorLike]]
|
distribution of interest, i.e., the distribution of interest is a |
OM[Inputs, Uniform[TensorLike]]
|
discrete, uniform distribution over the list of returned points. If |
OM[Inputs, Uniform[TensorLike]]
|
z is multi-input, returns a distribution for each input. |
Source code in trulens_explain/trulens/nn/distributions.py
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 |
|
__init__(cut=None)
¶
"Initialize DoI
Parameters:
Name | Type | Description | Default |
---|---|---|---|
cut |
Cut
|
The Cut in which the DoI will be applied. If |
None
|
Source code in trulens_explain/trulens/nn/distributions.py
49 50 51 52 53 54 55 56 57 58 |
|
cut()
¶
Returns:
Type | Description |
---|---|
Cut
|
The Cut in which the DoI will be applied. If |
Cut
|
applied to the input. otherwise, the distribution should be applied |
Cut
|
to the latent space defined by the cut. |
Source code in trulens_explain/trulens/nn/distributions.py
125 126 127 128 129 130 131 132 |
|
get_activation_multiplier(activation, *, model_inputs=None)
¶
Returns a term to multiply the gradient by to convert from "influence space" to "attribution space". Conceptually, "influence space" corresponds to the potential effect of a slight increase in each feature, while "attribution space" corresponds to an approximation of the net marginal contribution to the quantity of interest of each feature.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
activation |
OM[Inputs, TensorLike]
|
The activation of the layer the DoI is applied to. DoI may be multi-input in which case activation will be a list. |
required |
model_inputs |
Optional[ModelInputs]
|
Optional wrapped model input arguments that produce activation at cut. |
None
|
Returns:
Type | Description |
---|---|
OM[Inputs, TensorLike]
|
An array with the same shape as |
OM[Inputs, TensorLike]
|
multiplied by the gradient to obtain the attribution. The default |
OM[Inputs, TensorLike]
|
implementation of this method simply returns |
OM[Inputs, TensorLike]
|
activation is multi-input, returns one multiplier for each. |
Source code in trulens_explain/trulens/nn/distributions.py
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 |
|
DoiCutSupportError
¶
Bases: ValueError
Exception raised if the distribution of interest is called on a cut whose output is not supported by the distribution of interest.
Source code in trulens_explain/trulens/nn/distributions.py
34 35 36 37 38 39 |
|
GaussianDoi
¶
Bases: DoI
Distribution representing a Gaussian ball around the point. Used by Smooth Gradients.
Source code in trulens_explain/trulens/nn/distributions.py
420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 |
|
__init__(var, resolution, cut=None)
¶
Parameters:
Name | Type | Description | Default |
---|---|---|---|
var |
float
|
The variance of the Gaussian noise to be added around the point. |
required |
resolution |
int
|
Number of samples returned by each call to this DoI. |
required |
cut |
Cut
|
The Cut in which the DoI will be applied. If |
None
|
Source code in trulens_explain/trulens/nn/distributions.py
426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 |
|
LinearDoi
¶
Bases: DoI
Distribution representing the linear interpolation between a baseline and the given point. Used by Integrated Gradients.
Source code in trulens_explain/trulens/nn/distributions.py
241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 |
|
__init__(baseline=None, resolution=10, *, cut=None)
¶
The DoI for point, z
, will be a uniform distribution over the points
on the line segment connecting z
to baseline
, approximated by a
sample of resolution
points equally spaced along this segment.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
cut |
Cut, optional, from DoI
|
The Cut in which the DoI will be applied. If |
None
|
baseline |
BaselineLike
|
The baseline to interpolate from. Must be same shape as the
space the distribution acts over, i.e., the shape of the points,
|
None
|
resolution |
int
|
Number of points returned by each call to this DoI. A higher resolution is more computationally expensive, but gives a better approximation of the DoI this object mathematically represents. |
10
|
Source code in trulens_explain/trulens/nn/distributions.py
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 |
|
get_activation_multiplier(activation, *, model_inputs=None)
¶
Returns a term to multiply the gradient by to convert from "influence space" to "attribution space". Conceptually, "influence space" corresponds to the potential effect of a slight increase in each feature, while "attribution space" corresponds to an approximation of the net marginal contribution to the quantity of interest of each feature.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
activation |
OM[Inputs, TensorLike]
|
The activation of the layer the DoI is applied to. |
required |
Returns:
Type | Description |
---|---|
Inputs[TensorLike]
|
The activation adjusted by the baseline passed to the constructor. |
Source code in trulens_explain/trulens/nn/distributions.py
333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 |
|
PointDoi
¶
Bases: DoI
Distribution that puts all probability mass on a single point.
Source code in trulens_explain/trulens/nn/distributions.py
213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 |
|
__init__(cut=None)
¶
"Initialize PointDoI
Parameters:
Name | Type | Description | Default |
---|---|---|---|
cut |
Cut
|
The Cut in which the DoI will be applied. If |
None
|
Source code in trulens_explain/trulens/nn/distributions.py
218 219 220 221 222 223 224 225 226 227 |
|