refactor: change comments for clarity

Author: Michal-Novomestsky

pymc_extras/inference/laplace.py

@@ -106,7 +106,7 @@ def get_conditional_gaussian_approximation(
     jac = pytensor.gradient.grad(f_x, x)
     hess = pytensor.gradient.jacobian(jac.flatten(), x)
 
-    # Component of log(p(x | y, params)) which depends on x
+    # log(p(x | y, params)) only including terms that depend on x for the minimization step (logdet(Q) ignored as it is a constant wrt x)
     log_x_posterior = f_x - 0.5 * (x - mu).T @ Q @ (x - mu)
 
     # Maximize log(p(x | y, params)) wrt x
@@ -129,7 +129,8 @@ def get_conditional_gaussian_approximation(
         -0.5 * x.T @ (-hess + Q) @ x + x.T @ (Q @ mu + jac - hess @ x0) + 0.5 * logdetQ
     )
 
-    # TODO Currently x being passed in as an initial guess for x0 AND then also going to the true value of x
+    # Currently x is passed both as the query point for f(x, args) = logp(x | y, params) AND as an initial guess for x0. This may cause issues if the query point is
+    # far from the mode x0 or in a neighbourhood which results in poor convergence.
     return pytensor.function(args, [x0, conditional_gaussian_approx])