search

wujian16 / Cornell-MOE

A Python library for the state-of-the-art Bayesian optimization algorithms, with the core implemented in C++.
Other
261 stars 64 forks source link

the range of search domain #68

Open zhangyunzhao opened 5 years ago

zhangyunzhao commented 5 years ago

when i set the search_domain of learning_rate ,it can not work.

class CIFAR10(object):
    def __init__(self):
        self._dim = 1
        self._search_domain = numpy.array([[0.00001, 0.002]])
        self._num_init_pts = 8
        self._sample_var = 0.0
        self._min_value = 0.0
        self._observations = []
        self._num_fidelity = 0

    def train(self, x):
        try:
            # Data loading and preprocessing
            # The data, shuffled and split between train and test sets:
            num_classes = 10
            # The data, shuffled and split between train and test sets:
            (x_train, y_train), (x_test, y_test) = cifar10.load_data()

            # Convert class vectors to binary class matrices.
            y_train = keras.utils.to_categorical(y_train, num_classes)
            y_test = keras.utils.to_categorical(y_test, num_classes)

            lr = x
            lr = lr
            unit1 = int(pow(2, 5))
            unit2 = int(pow(2, 5))
            unit3 = int(pow(2, 5))
            batch_size = int(256)

cholesky matrix singular -8.577649091442408837E-06 Thread 3 of 8 failed on iteration 8 of 20. Message: SingularMatrixException: 8 x 8 matrix is singular; 7-th leading minor is not SPD. GP-Variance matrix singular. Check for duplicate points_to_sample/being_sampled or points_to_sample/being_sampled duplicating points_sampled with 0 noise. void optimal_learning::ExpectedImprovementEvaluator::ComputeGradExpectedImprovement(optimal_learning::ExpectedImprovementEvaluator::StateType, double) const (/home/test_data/zhaozhangyun/Cornell-MOE-master/moe/optimal_learning/cpp/gpp_math.cpp: 2066) cholesky matrix singular -4.816341968136730829E-05 Thread 5 of 8 failed on iteration 14 of 20. Message: SingularMatrixException: 8 x 8 matrix is singular; 8-th leading minor is not SPD. GP-Variance matrix singular. Check for duplicate points_to_sample/being_sampled or points_to_sample/being_sampled duplicating points_sampled with 0 noise. void optimal_learning::ExpectedImprovementEvaluator::ComputeGradExpectedImprovement(optimal_learning::ExpectedImprovementEvaluator::StateType, double) const (/home/test_data/zhaozhangyun/Cornell-MOE-master/moe/optimal_learning/cpp/gpp_math.cpp: 2066) Traceback (most recent call last): File "main.py", line 183, in cpp_sgd_params_kg, num_to_sample, num_mc=2 * 10) File "/home/test_data/zhaozhangyun/Cornell-MOE-master/examples/bayesian_optimization.py", line 31, in gen_sample_from_qei max_num_threads=8)) File "/home/test_data/auto_ml/venv2.7/lib/python2.7/site-packages/moe/optimal_learning/python/cpp_wrappers/expected_improvement.py", line 101, in multistart_expected_improvement_optimization status, moe.build.GPP.SingularMatrixException: SingularMatrixException: 8 x 8 matrix is singular; 7-th leading minor is not SPD. GP-Variance matrix singular. Check for duplicate points_to_sample/being_sampled or points_to_sample/being_sampled duplicating points_sampled with 0 noise. void optimal_learning::ExpectedImprovementEvaluator::ComputeGradExpectedImprovement(optimal_learning::ExpectedImprovementEvaluator::StateType, double*) const (/home/test_data/zhaozhangyun/Cornell-MOE-master/moe/optimal_learning/cpp/gpp_math.cpp: 2066)