DPR

Actor：

language model

$$\begin{equation} P(a_t | a_0, \dots, a_{t-1}) = P(a_t | c_t) \end{equation}$$

category

plan2vec

plan sequences classification

观测序列

$$\begin{equation} \bar{o} = \sum \limits_{i=1}^n \phi(o_i) \end{equation}$$
根据 $\bar{o}$确定最相似的plan的类别，确定观测序列所属的主题 $T$，然后在主题$T$中选出最相似的 $k$个plan

$$\begin{equation} \bar{s} = TopKSim(p_{ij}, \bar{o})\\ \tilde{s} = \sum_i \alpha_i s_i\\ s_i \in \bar{s}\\ \alpha_i = p_i \odot \bar{o}\\ \tilde{o_t} = f(\bar{o}, a_t)\\ s_t = concatenate(\tilde{o_t}, \tilde{s_t}) \end{equation}$$

$E[R_{1:\infty}]$