bayes_op&lightGBM

[JiHong1235676]

[LightGBM] [Warning] Using self-defined objective function
Traceback (most recent call last):
File "D:\Conda\anzhuang\lib\site-packages\bayes_opt\target_space.py", line 191, in probe
target = self._cache[_hashable(x)]
KeyError: (0.4224347626508457,)

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "E:\During_DR\PYradar\1_paper\git_bys.py", line 118, in
LGB_BO.maximize(init_points=2, n_iter=2)
File "D:\Conda\anzhuang\lib\site-packages\bayes_opt\bayesian_optimization.py", line 185, in maximize
self.probe(x_probe, lazy=False)
File "D:\Conda\anzhuang\lib\site-packages\bayes_opt\bayesian_optimization.py", line 116, in probe
self._space.probe(params)
File "D:\Conda\anzhuang\lib\site-packages\bayes_opt\target_space.py", line 194, in probe
target = self.target_func(**params)
File "E:\During_DR\PYradar\1_paper\git_bys.py", line 110, in LGB_CV
return -MSE(pf, PF)
File "D:\Conda\anzhuang\lib\site-packages\sklearn\utils\validation.py", line 63, in inner_f
return f(*args, **kwargs)
File "D:\Conda\anzhuang\lib\site-packages\sklearn\metrics_regression.py", line 335, in mean_squared_error
y_type, y_true, y_pred, multioutput = _check_reg_targets(
File "D:\Conda\anzhuang\lib\site-packages\sklearn\metrics_regression.py", line 88, in _check_reg_targets
check_consistent_length(y_true, y_pred)
File "D:\Conda\anzhuang\lib\site-packages\sklearn\utils\validation.py", line 316, in check_consistent_length
lengths = [_num_samples(X) for X in arrays if X is not None]
File "D:\Conda\anzhuang\lib\site-packages\sklearn\utils\validation.py", line 316, in
lengths = [_num_samples(X) for X in arrays if X is not None]
File "D:\Conda\anzhuang\lib\site-packages\sklearn\utils\validation.py", line 259, in _num_samples
raise TypeError("Singleton array %r cannot be considered"
TypeError: Singleton array 0.15225424981522542 cannot be considered a valid collection.

[JiHong1235676]

This is code
`def LGB_CV(alpha_0):
PF = 1
focal_loss = lambda x, y: focal_loss_lgb(x, y, alpha_0, 0) ###0.7
eval_error = lambda x, y: focal_loss_lgb_eval_error(x, y, alpha_0, 0) ##3
lgtrain = lgb.Dataset(
Xtrain, Ytrain,
free_raw_data=False)
params = {
'task': 'train',
'boosting_type': 'gbdt',
'objective': 'binary',
# 'objective': 'regression',
'metric': {'binary_logloss'},
'max_depth': 11,
'num_leaves': 25,
'num_trees': 80,
'learning_rate': 0.2,
# 'num_threads':40,
'feature_fraction': 0.9,
'bagging_fraction': 0.9,
'bagging_freq': 5,
'verbose': 0
}
t0 = time.time()
model = lgb.train(
params,
lgtrain,
# valid_sets=[lgvalid],
fobj=focal_loss,
feval=eval_error
)
preds = model.predict(Xtrain, num_iteration=model.best_iteration)
def sigmoid(x): return 1. / (1. + np.exp(-x))
preds = sigmoid(preds)
y_pred = [int(p > 0.5) for p in preds]
confusion_matrix1 = confusion_matrix(Ytrain, y_pred) # ,labels=[0,1]
TP = confusion_matrix1[1, 1]
FP = confusion_matrix1[0, 1]
pf = FP / (TP + FP)

return -MSE(pf, PF)

pbounds={'alpha_0':(0.01,0.999)}
from bayes_opt import BayesianOptimization
LGB_BO = BayesianOptimization(LGB_CV,
pbounds=pbounds,
random_state=1,)
LGB_BO.maximize(init_points=2, n_iter=2)
print(LGB_BO.max)`

[till-m]

Hi @JiHong1235676,

I can't run your code because a few things are missing (imports, etc). Please also try to wrap long codeblocks with triple backticks ``` so that indentation is preserved and the code remains readable.
By just analyzing the stack trace, I think that the problem occurring here does not stem from this package. Rather, the error seems to be produced by one of the arguments of MSE being a singleton array.

This means that an np.ndarray was generated like this

a = np.array(3.141)

and not wrapped in an iterable

a = np.array([3.141])

something you could try is to replace the line return -MSE(pf, PF) with

return -MSE(np.atleast_1d(pf), np.atleast_1d(PF))

I hope this helps. Let me know if it works.

[JiHong1235676]

Thank you very much, and I'll debug my code this way. Since the data I use is private, it is not convenient to upload.

[JiHong1235676]

Hi @JiHong1235676,

I can't run your code because a few things are missing (imports, etc). Please also try to wrap long codeblocks with triple backticks ``` so that indentation is preserved and the code remains readable. By just analyzing the stack trace, I think that the problem occurring here does not stem from this package. Rather, the error seems to be produced by one of the arguments of MSE being a singleton array.

This means that an np.ndarray was generated like this

a = np.array(3.141)

and not wrapped in an iterable

a = np.array([3.141])

something you could try is to replace the line return -MSE(pf, PF) with

return -MSE(np.atleast_1d(pf), np.atleast_1d(PF))

I hope this helps. Let me know if it works.

Thank you for solving my problem! I changed the MSE function into subtraction. The code runs successfully.

[till-m]

Since the data I use is private, it is not convenient to upload

I understand, ideally in this case you would provide toy data.

Thank you for solving my problem!

No problem, feel free to open another issue in case you run into trouble again :)

@bwheelz36 I think this can be closed.