A Appendix Sum Product Networks Causal Inference with Tractable Probabilistic Models

In the main paper we have mostly considered perfect interventions i.e., interventions that render the However, as already suggested, our model is not restricted to any specific intervention type or instantiation. Nonetheless, it can be observed that some interventions are being modelled more precisely than others, e.g. A.2 On the Visualization of Mean Densities For Earthquake and Cancer data sets, we use 5 different number of sum node weights: 600, 1200, 1800, 2400 and 3200. For the synthetic causal health data set we use 300, 600, 1000, 1500, 2000. The results are overall consistent with the presented i.e., the mean performance are consistent across Table 2: Dimensions of the used data sets .