Skip to content

Alg setup

Provides methods for the ATHENA project for setting DEAP structures used in software and functions used by DEAP during evolutionary algorithm.

Setup of DEAP structures and additional functions used in GENN algorithm

configure_toolbox(fitness, selection, crosstype='match', init='sensible')

Configure the DEAP toolbox for controlling GE algorithm

Parameters:

Name Type Description Default
fitness str

SNP values filename

 required
selection str

type of selection operator

 required
crosstype str

type of crossover operator

 'match'
init str

scale outcome values from 0 to 1.0

 'sensible'

Returns:

Type Description
Toolbox

DEAP base.Toolbox configured for a GE run
Source code in src/athenage/genn/alg_setup.py 
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
def configure_toolbox(fitness: str, selection: str, crosstype:str ='match',
                      init:str ='sensible') -> base.Toolbox:
    """Configure the DEAP toolbox for controlling GE algorithm

    Args:
        fitness: SNP values filename
        selection: type of selection operator
        crosstype: type of crossover operator
        init: scale outcome values from 0 to 1.0

    Returns:
        DEAP base.Toolbox configured for a GE run
    """

    toolbox = base.Toolbox()
    creator.create("FitnessMax", base.Fitness, weights=(1.0,))
    creator.create('Individual', grape.Individual, fitness=creator.FitnessMax)
    if init == 'sensible':
        toolbox.register("populationCreator", grape.sensible_initialization, creator.Individual) 
    else:
        toolbox.register("populationCreator", grape.random_initialization, creator.Individual) 

    if crosstype == 'onepoint':
        toolbox.register("mate", grape.crossover_onepoint)
    elif crosstype == 'match':
        toolbox.register("mate", grape.crossover_match)
    elif crosstype == 'block':
        toolbox.register("mate", grape.crossover_block)

    toolbox.register("mutate", grape.mutation_int_flip_per_codon)

    if fitness=='r-squared':
        if selection == 'epsilon_lexicase':
            toolbox.register("evaluate", fitness_rsquared_lexicase)
            toolbox.register("select", grape.selAutoEpsilonLexicase)#, tournsize=7)
        else:
            toolbox.register("evaluate", fitness_rsquared)
            toolbox.register("select", tools.selTournament, tournsize=2)
    elif fitness == 'balanced_acc':
        if selection=='lexicase':
            toolbox.register("evaluate", fitness_balacc_lexicase)
            toolbox.register("select", grape.selBalAccLexicase)
        else:
            toolbox.register("evaluate", fitness_balacc)
            toolbox.register("select", tools.selTournament, tournsize=2)
    elif fitness == 'auc':
        if selection=='lexicase':
            toolbox.register("evaluate", fitness_auc_lexicase)
            toolbox.register("select", grape.selBalAccLexicase)
        else:
            toolbox.register("evaluate", fitness_auc)
            toolbox.register("select", tools.selTournament, tournsize=2)
    else:
        raise ValueError("fitness must be fitness_rsquared or fitness_balacc")

    return toolbox

fitness_auc(individual, points)

Calculate area under the curve (AUC) as fitness for this individual using points passed

Parameters:

Name Type Description Default
individual Individual

solution being evaluated for fitness

 required
points list

2-D list containing inputs and outcome for calculating fitness points[0] contains 2-D np.ndarray of all inputs

 required

Returns:

Type Description
float

AUC fitness
Source code in src/athenage/genn/alg_setup.py 
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
def fitness_auc(individual: 'deap.creator.Individual', points: list) -> float:
    """Calculate area under the curve (AUC)
    as fitness for this individual using points passed

    Args:
        individual: solution being evaluated for fitness
        points: 2-D list containing inputs and outcome for calculating fitness
            points[0] contains 2-D np.ndarray of all inputs

    Returns:
        AUC fitness
    """

    x = points[0]
    y = points[1]


    if individual.invalid == True:
        return INVALID_FITNESS,

    try:
        pred = eval(individual.phenotype)

#         pred2 = eval(compress_weights(individual.phenotype))
    except (FloatingPointError, ZeroDivisionError, OverflowError,
            MemoryError, ValueError):
        return INVALID_FITNESS,
    except Exception as err:
            # Other errors should not usually happen (unless we have
            # an unprotected operator) so user would prefer to see them.
            print("evaluation error", err)
            raise
    assert np.isrealobj(pred)

    try:
        nan_mask = np.isnan(pred)
        # assign case/control status
        pred_nonan = np.where(pred[~nan_mask] < 0.5, 0, 1)
        fitness = roc_auc_score(y[~nan_mask],pred_nonan)
        individual.nmissing = np.count_nonzero(np.isnan(pred))

#         fitness_compressed = balanced_accuracy_score(y[~nan_mask],pred2)
    except (FloatingPointError, ZeroDivisionError, OverflowError,
            MemoryError, ValueError):
        fitness = INVALID_FITNESS
    except Exception as err:
            # Other errors should not usually happen (unless we have
            # an unprotected operator) so user would prefer to see them.
            print("fitness error", err)
            raise

    if fitness == float("inf"):
        return INVALID_FITNESS,

    return fitness,

fitness_auc_lexicase(individual, points)

Calculate area under the curve (AUC) for this individual and store differences in predicted vs observed outcomes for use in lexicase selection

Parameters:

Name Type Description Default
individual Individual

solution being evaluated for fitness

 required
points list

2-D list containing inputs and outcome for calculating fitness points[0] contains 2-D np.ndarray of all inputs

 required

Returns:

Type Description
float

balanced accuracy fitness
Source code in src/athenage/genn/alg_setup.py 
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
def fitness_auc_lexicase(individual: 'deap.creator.Individual', points: list) -> float:
    """Calculate area under the curve (AUC) for this individual and store differences in
        predicted vs observed outcomes for use in lexicase selection

    Args:
        individual: solution being evaluated for fitness
        points: 2-D list containing inputs and outcome for calculating fitness
            points[0] contains 2-D np.ndarray of all inputs

    Returns:
        balanced accuracy fitness
    """

    x = points[0]
    y = points[1]

    if individual.invalid == True:
        individual.ptscores = np.full(len(y), np.nan)
        return INVALID_FITNESS,

    try:
        pred = eval(individual.phenotype)
    except (FloatingPointError, ZeroDivisionError, OverflowError,
            MemoryError, ValueError):
        return INVALID_FITNESS,
    except Exception as err:
            # Other errors should not usually happen (unless we have
            # an unprotected operator) so user would prefer to see them.
            print("evaluation error", err)
            raise
    assert np.isrealobj(pred)

    try:
        nan_mask = np.isnan(pred)
        # assign case/control status
        pred_nonan = np.where(pred[~nan_mask] < 0.5, 0, 1)
        fitness = roc_auc_score(y[~nan_mask],pred_nonan)
        individual.nmissing = np.count_nonzero(np.isnan(pred))

        # save individual point scores for use in lexicase selection
        full = np.copy(pred)
        full[~nan_mask] = np.where(pred[~nan_mask] < 0.5, 0, 1)
        individual.ptscores = np.absolute(y-full)


    except (FloatingPointError, ZeroDivisionError, OverflowError,
            MemoryError, ValueError):
        fitness = INVALID_FITNESS
        individual.ptscores = np.full(len(y), np.nan)
    except Exception as err:
            # Other errors should not usually happen (unless we have
            # an unprotected operator) so user would prefer to see them.
            print("fitness error", err)
            raise

    if fitness == float("inf"):
        individual.ptscores = np.full(len(y), np.nan)
        return INVALID_FITNESS,

    return fitness,

fitness_balacc(individual, points)

Calculate balanced accuracy as fitness for this individual using points passed

Parameters:

Name Type Description Default
individual Individual

solution being evaluated for fitness

 required
points list

2-D list containing inputs and outcome for calculating fitness points[0] contains 2-D np.ndarray of all inputs

 required

Returns:

Type Description
float

balanced accuracy fitness
Source code in src/athenage/genn/alg_setup.py 
207
208
209
210
211
212
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
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
def fitness_balacc(individual: 'deap.creator.Individual', points: list) -> float:
    """Calculate balanced accuracy as fitness for this individual using points passed

    Args:
        individual: solution being evaluated for fitness
        points: 2-D list containing inputs and outcome for calculating fitness
            points[0] contains 2-D np.ndarray of all inputs

    Returns:
        balanced accuracy fitness
    """

    x = points[0]
    y = points[1]


    if individual.invalid == True:
        return INVALID_FITNESS,

    try:
        pred = eval(individual.phenotype)

#         pred2 = eval(compress_weights(individual.phenotype))
    except (FloatingPointError, ZeroDivisionError, OverflowError,
            MemoryError, ValueError):
        return INVALID_FITNESS,
    except Exception as err:
            # Other errors should not usually happen (unless we have
            # an unprotected operator) so user would prefer to see them.
            print("evaluation error", err)
            raise
    assert np.isrealobj(pred)

    try:
        nan_mask = np.isnan(pred)
        # assign case/control status
        pred_nonan = np.where(pred[~nan_mask] < 0.5, 0, 1)
        fitness = balanced_accuracy_score(y[~nan_mask],pred_nonan)
        individual.nmissing = np.count_nonzero(np.isnan(pred))

#         fitness_compressed = balanced_accuracy_score(y[~nan_mask],pred2)
    except (FloatingPointError, ZeroDivisionError, OverflowError,
            MemoryError, ValueError):
        fitness = INVALID_FITNESS
    except Exception as err:
            # Other errors should not usually happen (unless we have
            # an unprotected operator) so user would prefer to see them.
            print("fitness error", err)
            raise

    if fitness == float("inf"):
        return INVALID_FITNESS,

    return fitness,

fitness_balacc_lexicase(individual, points)

Calculate balanced accuracy fitness for this individual and store differences in predicted vs observed outcomes for use in lexicase selection

Parameters:

Name Type Description Default
individual Individual

solution being evaluated for fitness

 required
points list

2-D list containing inputs and outcome for calculating fitness points[0] contains 2-D np.ndarray of all inputs

 required

Returns:

Type Description
float

balanced accuracy fitness
Source code in src/athenage/genn/alg_setup.py 
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
def fitness_balacc_lexicase(individual: 'deap.creator.Individual', points: list) -> float:
    """Calculate balanced accuracy fitness for this individual and store differences in
        predicted vs observed outcomes for use in lexicase selection

    Args:
        individual: solution being evaluated for fitness
        points: 2-D list containing inputs and outcome for calculating fitness
            points[0] contains 2-D np.ndarray of all inputs

    Returns:
        balanced accuracy fitness
    """

    x = points[0]
    y = points[1]

    if individual.invalid == True:
        individual.ptscores = np.full(len(y), np.nan)
        return INVALID_FITNESS,

    try:
        pred = eval(individual.phenotype)
    except (FloatingPointError, ZeroDivisionError, OverflowError,
            MemoryError, ValueError):
        return INVALID_FITNESS,
    except Exception as err:
            # Other errors should not usually happen (unless we have
            # an unprotected operator) so user would prefer to see them.
            print("evaluation error", err)
            raise
    assert np.isrealobj(pred)

    try:
        nan_mask = np.isnan(pred)
        # assign case/control status
        pred_nonan = np.where(pred[~nan_mask] < 0.5, 0, 1)
        fitness = balanced_accuracy_score(y[~nan_mask],pred_nonan)
        individual.nmissing = np.count_nonzero(np.isnan(pred))

        # save individual point scores for use in lexicase selection
        full = np.copy(pred)
        full[~nan_mask] = np.where(pred[~nan_mask] < 0.5, 0, 1)
        individual.ptscores = np.absolute(y-full)


    except (FloatingPointError, ZeroDivisionError, OverflowError,
            MemoryError, ValueError):
        fitness = INVALID_FITNESS
        individual.ptscores = np.full(len(y), np.nan)
    except Exception as err:
            # Other errors should not usually happen (unless we have
            # an unprotected operator) so user would prefer to see them.
            print("fitness error", err)
            raise

    if fitness == float("inf"):
        individual.ptscores = np.full(len(y), np.nan)
        return INVALID_FITNESS,

    return fitness,

fitness_rsquared(individual, points)

Calculate r-squared fitness for this individual using points passed

Parameters:

Name Type Description Default
individual Individual

solution being evaluated for fitness

 required
points list

2-D list containing inputs and outcome for calculating fitness points[0] contains 2-D np.ndarray of all inputs

 required

Returns:

Type Description
float

r-squared fitness
Source code in src/athenage/genn/alg_setup.py 
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
def fitness_rsquared(individual: 'deap.creator.Individual', points: list) -> float:
    """Calculate r-squared fitness for this individual using points passed

    Args:
        individual: solution being evaluated for fitness
        points: 2-D list containing inputs and outcome for calculating fitness
            points[0] contains 2-D np.ndarray of all inputs

    Returns:
        r-squared fitness
    """
    x = points[0]
    y = points[1]

    if individual.invalid == True:
        return INVALID_FITNESS,

    try:
        pred = eval(individual.phenotype)
    except (FloatingPointError, ZeroDivisionError, OverflowError,
            MemoryError, ValueError):
        return INVALID_FITNESS,
    except Exception as err:
            # Other errors should not usually happen (unless we have
            # an unprotected operator) so user would prefer to see them.
            print("evaluation error", err)
            raise
    assert np.isrealobj(pred)

    try:
        fitness = r_squared(y,pred)
        individual.nmissing = np.count_nonzero(np.isnan(pred))
    except (FloatingPointError, ZeroDivisionError, OverflowError,
            MemoryError, ValueError):
        fitness = INVALID_FITNESS
    except Exception as err:
            # Other errors should not usually happen (unless we have
            # an unprotected operator) so user would prefer to see them.
            print("fitness error", err)
            raise

    if fitness == float("inf"):
        return INVALID_FITNESS,

    return fitness,

fitness_rsquared_lexicase(individual, points)

Calculate r-squared fitness for this individual and store differences in predicted vs observed outcomes for use in lexicase selection

Parameters:

Name Type Description Default
individual Individual

solution being evaluated for fitness

 required
points list

2-D list containing inputs and outcome for calculating fitness points[0] contains 2-D np.ndarray of all inputs

 required

Returns:

Type Description
float

r-squared fitness
Source code in src/athenage/genn/alg_setup.py 
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
def fitness_rsquared_lexicase(individual: 'deap.creator.Individual', points: list) -> float:
    """Calculate r-squared fitness for this individual and store differences in
        predicted vs observed outcomes for use in lexicase selection

    Args:
        individual: solution being evaluated for fitness
        points: 2-D list containing inputs and outcome for calculating fitness
            points[0] contains 2-D np.ndarray of all inputs

    Returns:
        r-squared fitness
    """

    x = points[0]
    y = points[1]

    if individual.invalid == True:
        individual.ptscores = np.full(len(y), np.nan)
        return INVALID_FITNESS,

    try:
        pred = eval(individual.phenotype)
    except (FloatingPointError, ZeroDivisionError, OverflowError,
            MemoryError, ValueError):
        return INVALID_FITNESS,
    except Exception as err:
            # Other errors should not usually happen (unless we have
            # an unprotected operator) so user would prefer to see them.
            print("evaluation error", err)
            raise
    assert np.isrealobj(pred)

    try:
        fitness = r_squared(y,pred)
        individual.nmissing = np.count_nonzero(np.isnan(pred))

        # store individual differences for lexicase
        individual.ptscores = np.absolute(y-pred)
    except (FloatingPointError, ZeroDivisionError, OverflowError,
            MemoryError, ValueError):
        individual.ptscores = np.full(len(y), np.nan)
        fitness = INVALID_FITNESS
    except Exception as err:
            # Other errors should not usually happen (unless we have
            # an unprotected operator) so user would prefer to see them.
            print("fitness error", err)
            raise

    if fitness == float("inf"):
        individual.ptscores = np.full(len(y), np.nan)
        return INVALID_FITNESS,

    return fitness,

r_squared(y, y_hat)

Calculate r-squared values

Parameters:

Name Type Description Default
y ndarray

Observed values

 required
y_hat ndarray

Predicted values

 required

Returns:

Type Description
float

r-squared value
Source code in src/athenage/genn/alg_setup.py 
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
def r_squared(y: np.ndarray, y_hat: np.ndarray) -> float:
    """Calculate r-squared values

    Args:
        y: Observed values
        y_hat: Predicted values

    Returns:
        r-squared value
    """

    nan_mask = np.isnan(y_hat)
    y_bar = y[~nan_mask].mean()
    ss_tot = ((y[~nan_mask]-y_bar)**2).sum()
    ss_res = ((y[~nan_mask]-y_hat[~nan_mask])**2).sum()
    return 1 - (ss_res/ss_tot)

set_crossover(toolbox, crosstype)

Sets crossover type for toolbox

Parameters:

Name Type Description Default
toolbox toolbox

DEAP toolbox

 required
crosstype str

specifies type to use

 required

Returns:

Type Description
None

None
Source code in src/athenage/genn/alg_setup.py 
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
def set_crossover(toolbox: 'deap.base.toolbox', crosstype: str) -> None:
    """Sets crossover type for toolbox

    Args:
        toolbox: DEAP toolbox
        crosstype: specifies type to use

    Returns:
        None
    """
    if crosstype == 'onepoint':
        toolbox.register("mate", grape.crossover_onepoint)
    elif crosstype == 'match':
        toolbox.register("mate", grape.crossover_match)
    elif crosstype == 'block':
        toolbox.register("mate", grape.crossover_block)