updates

Author: msneubauer

_sources/lectures/VariationalInference.ipynb

@@ -752,7 +752,9 @@
     " - Minimizing $-\\text{ELBO}(q)$ with respect to $q$ is equivalent to minimizing $\\text{KL}(q\\parallel p)$.\n",
     "\n",
     " - $\\text{ELBO}(q)$ is much easier to calculate since it does not depend on the evidence $P(D)$.\n",
-    " \n",
+    "\n",
+    "The ELBO is strongly connected to [information theory and entropy](https://en.wikipedia.org/wiki/Information_theory). Specifically, the ELBO is not just a loss function for variational inference; it is an information-theoretic objective that balances how much information from the input data is encoded into the latent space (bits) versus how accurately that data can be reconstructed. \n",
+    "\n",
     "Note that, as with the KL divergence, the ELBO can be evaluated in terms of expectation values,\n",
     "\n",
     "$$ \\Large\n",