In a bayesian network ,

the complexity of calculating conditional probability distributions necessary to express uncertain influences, is exponential 2 ^ N (N: number of network parameters: the number of nodes and their probabilities) this cmplexity increases rapidly with N (NP Hard problem), so it is essential to keep N under control.

but according to some reference this problem can be solved if we use simplification's methods of of the computation (by simplifying the graphs) like the Method of regrouping (Junction tree Algorithm).

my question :

is the Junction tree Algorithm sufficient to avoid the problem of exponential time compelexity even if the number of nodes is large?

Thank you in advance