Rephrase \beta description. (#831)

"More fluctate" is somewhat difficult to parse.

Co-authored-by: Patrick Collins <Patrick Collins>

Author: pscollins

docs/en/week07/07-1.md

@@ -82,7 +82,11 @@ We can look at the energies as unnormalised negative log probabilities, and use
 
 $$P(y \mid x) = \frac{\exp (-\beta F(x,y))}{\int_{y'}\exp(-\beta F(x,y'))}$$
 
-where $\beta$ is positive constant and needs to be calibrated to fit your model. Larger $\beta$ gives a more fluctuate model while smaller $\beta$ gives a smoother model. (In physics, $\beta$ is inverse temperature: $\beta \rightarrow \infty$ means temperature goes to zero).
+where $\beta$ is positive constant and needs to be calibrated to fit your model:
+as $\beta \rightarrow \infty$, this function converges to the argmax function;
+smaller values of $\beta$ leads to a smoother distribution.  (In physics,
+$\beta$ is inverse temperature: $\beta \rightarrow \infty$ means temperature
+goes to zero).
 
 $$P(y,z \mid x) = \frac{\exp(-\beta F(x,y,z))}{\int_{y}\int_{z}\exp(-\beta F(x,y,z))}$$