Calculating the Log Partition Function in LBP about pgmax HOT 2 CLOSED

Given infer.compute_energy to compute the energy $E(x)$ for a state $x$ in a graphical model, the probability of a state $x$ using the Boltzmann distribution can be expressed as:

$$P(x) = \frac{\exp(-E(x))}{Z}$$

where $Z$ is the partition function, and $E(x)$ is the energy of state $x$. The partition function $Z$ is the sum over all possible states:

$$Z = \sum_{x_i} \exp(-E(x_i))$$

To find the logarithm of the partition function, $\log Z$, use the log-sum-exp (LSE) function over the negative energies of all states:
$$\log Z = \log \left( \sum_{x_i} \exp(-E(x_i)) \right) = \text{LSE}(-E(x_i))$$

would that be an appropriate way to compute $\log Z$?

from pgmax.