mindscience.common.batched_hessian
- mindscience.common.batched_hessian(model)[源代码]
计算网络模型的海森矩阵。
说明
本函数在实现中使用了 mindspore.jacrev 接口来计算 Hessian 矩阵,因此要求 MindSpore 版本 >= 2.0.0。
- 参数:
model (mindspore.nn.Cell) - 输入维度为 in_channels 输出维度为 out_channels 的网络模型。
- 返回:
Tensor,用于计算海森矩阵的 Hessian 实例。输入维度为 \([batch_size, in_channels]\) ,输出维度为 \([out_channels, in_channels, batch_size, in_channels]\)。
样例:
>>> import numpy as np >>> from mindspore import nn, ops, Tensor >>> from mindspore import dtype as mstype >>> from mindscience.common import batched_hessian >>> np.random.seed(123456) >>> class Net(nn.Cell): ... def __init__(self, cin=2, cout=1, hidden=10): ... super().__init__() ... self.fc1 = nn.Dense(cin, hidden) ... self.fc2 = nn.Dense(hidden, hidden) ... self.fcout = nn.Dense(hidden, cout) ... self.act = ops.Tanh() ... ... def construct(self, x): ... x = self.act(self.fc1(x)) ... x = self.act(self.fc2(x)) ... x = self.fcout(x) ... return x >>> model = Net() >>> hessian = batched_hessian(model) >>> inputs = np.random.random(size=(3, 2)) >>> res = hessian(Tensor(inputs, mstype.float32)) >>> print(res.shape) (1, 2, 3, 2)