将纸张的茶色损失函数（平均负对数偏似性）从纸张转换为喀拉斯语

Question

我正在尝试将DeepSurv（https://bmcmedresmethodol.biomedcentral.com/articles/10.1186/s12874-018-0482-1）theano代码（https://github.com/jaredleekatzman/DeepSurv/blob/master/deepsurv/deep_surv.py）转换为keras格式。当前，我正在尝试创建成本/损失功能。

为了使人们不必在github页面上进行搜索，他们的代码是

def _negative_log_likelihood(self, E, deterministic = False):
        """Return the negative average log-likelihood of the prediction
            of this model under a given target distribution.
        .. math::
            \frac{1}{N_D} \sum_{i \in D}[F(x_i,\theta) - log(\sum_{j \in R_i} e^F(x_j,\theta))]
                - \lambda P(\theta)
        where:
            D is the set of observed events
            N_D is the number of observed events
            R_i is the set of examples that are still alive at time of death t_j
            F(x,\theta) = log hazard rate
        Note: We assume that there are no tied event times
        Parameters:
            E (n,): TensorVector that corresponds to a vector that gives the censor
                variable for each example
            deterministic: True or False. Determines if the output of the network
                is calculated deterministically.
        Returns:
            neg_likelihood: Theano expression that computes negative
                partial Cox likelihood
        """
        risk = self.risk(deterministic)
        hazard_ratio = T.exp(risk)
        log_risk = T.log(T.extra_ops.cumsum(hazard_ratio))
        uncensored_likelihood = risk.T - log_risk
        censored_likelihood = uncensored_likelihood * E
        num_observed_events = T.sum(E)
        neg_likelihood = -T.sum(censored_likelihood) / num_observed_events
        return neg_likelihood

因为我仍然是新的编码人员，所以我试图不必重写用于计算对数风险函数的函数，并且一直在探索生命线（https://lifelines.readthedocs.io/en/latest/）。

我的模型的 theta（xi）将是使用线性激活的神经网络的输出。

我最困惑的地方是

我打算在这里使用“生命线”。

from lifelines import CoxPHFitter
cph = CoxPHFitter()
cph.fit(x_norm_y_joined, duration_col='survival_fu_combine', event_col='death', show_progress=False)
cph.summary

test3 = pd.DataFrame(cph.summary).T

test3

然后我要用代码对exp（coef）行求和

test3.iloc[1,].sum()

最后得到

我会放

-np.log(test3.iloc[1,].sum())

我的思维过程正确吗？我有疑问，因为我不确定如何将零件编码为红色

E是

E (n,): TensorVector that corresponds to a vector that gives the censor
            variable for each example