Get the latest tech news
Computing Tricky Probabilities Using Model Counting
f certain events are really hard to estimate sometimes. Often, it’s because we lack information of the underlying causal chains, but sometimes, it’s because the causes are so intertwined that even if we know the underlying probabilities of certain events happening along with the causal chains, it’s still impossible for us to untangle the web.
Which is a straightforward 1-to-1 translation of the above, except we have to tell the counter that this is a projected weighted model counting problem (“pwmc”) and that it has 3 variables, and 2 constraints (“p cnf 3 2”). For example, if you want to run a hazardous process such as a nuclear reactor or a chemical plant, or you want to provide insurance for some complicated (financial) product, accurately computing the probabilities of bad events happening is extremely important. Similarly, if you want to bet for/against market moves, where you know or have a good idea of the underlying causal chains and associated probabilities, you can build quant trading, or expert advice systems based on the above.
Or read this on Hacker News