kalmannet

数据集的格式

​Simulations/Linear_CA/data/decimated_dt1e-2_T100_r0_randnInit.pt​

​​

保存这段数据的代码：

torch.save([train_input, train_target, cv_input, cv_target, test_input, test_target,train_init, cv_init, test_init], fileName)
​

0. train_input:train_batch_size,self.H.size()[0],T_MAX——记录观测值
1. train_target:train_batch_size,self.F.size()[0],T_MAX——记录状态
2. cv_input:cv_batch_size,self.H.size()[0],T_MAX
3. cv_target:cv_batch_size,self.F.size()[0],T_MAX
4. test_input:cv_batch_size,self.H.size()[0],T_MAX
5. test_target:cv_batch_size,self.F.size()[0],T_MAX
6. train_init:train_batch_size,self.F.size()[0],1——初始状态
7. cv_init:cv_batch_size,self.F.size()[0],1
8. test_init:cv_batch_size,self.F.size()[0],1

数据的再转换

构造真实状态对应观测的数据集，然后一些计算在 Kalmannet 里进行。

    def step_KGain_est(self, y):
        # both in size [batch_size, n]
        obs_diff = torch.squeeze(y,2) - torch.squeeze(self.y_previous,2) 
        obs_innov_diff = torch.squeeze(y,2) - torch.squeeze(self.m1y,2)
        # both in size [batch_size, m]
        fw_evol_diff = torch.squeeze(self.m1x_posterior,2) - torch.squeeze(self.m1x_posterior_previous,2)
        fw_update_diff = torch.squeeze(self.m1x_posterior,2) - torch.squeeze(self.m1x_prior_previous,2)

        obs_diff = func.normalize(obs_diff, p=2, dim=1, eps=1e-12, out=None)
        obs_innov_diff = func.normalize(obs_innov_diff, p=2, dim=1, eps=1e-12, out=None)
        fw_evol_diff = func.normalize(fw_evol_diff, p=2, dim=1, eps=1e-12, out=None)
        fw_update_diff = func.normalize(fw_update_diff, p=2, dim=1, eps=1e-12, out=None)

        # Kalman Gain Network Step
        KG = self.KGain_step(obs_diff, obs_innov_diff, fw_evol_diff, fw_update_diff)

        # Reshape Kalman Gain to a Matrix
        self.KGain = torch.reshape(KG, (self.batch_size, self.m, self.n))
​

‍