跳到正文

抽象先验

smac.model.gaussian_process.priors.abstract_prior #

AbstractPrior #

AbstractPrior(seed: int = 0)

定义高斯过程超参数先验接口的抽象基类。

此类改编自 RoBO

Klein, A. and Falkner, S. and Mansur, N. and Hutter, F. RoBO: A Flexible and Robust Bayesian Optimization Framework in Python In: NIPS 2017 Bayesian Optimization Workshop

注意#

当计算标量输入的 lnprob 或梯度时，我们使用 math.* 而不是 np.*。

参数#

seed : int, 默认为 0

源代码位于 smac/model/gaussian_process/priors/abstract_prior.py 
30
31
32
def __init__(self, seed: int = 0):
    self._seed = seed
    self._rng = np.random.RandomState(seed)

meta property #

meta: dict[str, Any]

返回创建对象的元数据。

get_gradient #

get_gradient(theta: float) -> float

计算先验关于 theta 的梯度。此方法内部调用 self._get_gradient。

警告#

Theta 必须是原始比例。

参数#

theta : float 对数空间的超参数配置

返回#

gradient : float 先验在 theta 处的梯度。

源代码位于 smac/model/gaussian_process/priors/abstract_prior.py 
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
def get_gradient(self, theta: float) -> float:
    """Computes the gradient of the prior with respect to theta. Internally, his method calls `self._get_gradient`.

    Warning
    -------
    Theta must be on the original scale.

    Parameters
    ----------
    theta : float
        Hyperparameter configuration in log space

    Returns
    -------
    gradient : float
        The gradient of the prior at theta.
    """
    return self._get_gradient(np.exp(theta))

get_log_probability #

get_log_probability(theta: float) -> float

返回 theta 的对数概率。此方法对 theta 取指数并调用 self._get_log_probability。

警告#

Theta 必须是对数比例！

参数#

theta : float 对数空间的超参数配置。

返回#

float theta 的对数概率

源代码位于 smac/model/gaussian_process/priors/abstract_prior.py 
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
def get_log_probability(self, theta: float) -> float:
    """Returns the log probability of theta. This method exponentiates theta and calls `self._get_log_probability`.

    Warning
    -------
    Theta must be on a log scale!

    Parameters
    ----------
    theta : float
        Hyperparameter configuration in log space.

    Returns
    -------
    float
        The log probability of theta
    """
    return self._get_log_probability(np.exp(theta))

sample_from_prior #

sample_from_prior(n_samples: int) -> ndarray

从先验中返回 n_samples 个样本。所有样本都在对数比例上。此方法调用 self._sample_from_prior 并对获得的值应用对数变换。

参数#

n_samples : int 将要抽取的样本数量。

返回#

samples : np.ndarray

源代码位于 smac/model/gaussian_process/priors/abstract_prior.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
def sample_from_prior(self, n_samples: int) -> np.ndarray:
    """Returns `n_samples` from the prior. All samples are on a log scale. This method calls
    `self._sample_from_prior` and applies a log transformation to the obtained values.

    Parameters
    ----------
    n_samples : int
        The number of samples that will be drawn.

    Returns
    -------
    samples : np.ndarray
    """
    if np.ndim(n_samples) != 0:
        raise ValueError("argument n_samples needs to be a scalar (is %s)" % n_samples)

    if n_samples <= 0:
        raise ValueError("argument n_samples needs to be positive (is %d)" % n_samples)

    sample = np.log(self._sample_from_prior(n_samples=n_samples))

    if np.any(~np.isfinite(sample)):
        raise ValueError("Sample %s from prior %s contains infinite values!" % (sample, self))

    return sample