Bug using pearl_inf_engine
Opened this issue · 0 comments
Deleted user commented
At some size of the problem, the estimation of probability for variables with 2 states was 0 and 0.
We now use gibbs_sampling_inf_engine.
The code is available on Figshare (https://figshare.com/articles/New_draft_itemA_modeling_approach_to_evaluate_the_influence_of_the_spatial_and_temporal_size_of_an_epidemiological_surveillance_network_on_phytosanitary_treatments_for_crops/4675759).
For example, in SIMULATE.m with n=196, b=2, it was OK until t=5, afterward all the probabilities was estimated at 0.
MJ Cros