Supplementary Material of PLANS: Neuro-Symbolic Program Learning from Videos

We provide the following appendices In A, we give additional information about the datasets (Karel and ViZDoom). In B, we describe precisely the neural component of PLANS and its training process. In C, we present the implementation of the rule-based solver. In D, we analyse the temporal complexity of PLANS. Table 3 contains high-level information about both datasets.