diff --git a/exercise_2_es/problem_ackley.ipynb b/exercise_2_es/problem_ackley.ipynb
index 40bf1e0..c003f45 100644
--- a/exercise_2_es/problem_ackley.ipynb
+++ b/exercise_2_es/problem_ackley.ipynb
@@ -194,7 +194,10 @@
     "3. How does a large $\\mu$/$\\lambda$ ratio influence the convergence rate?\n",
     "4. We can also introduce a learning schedule by a decaying mutation rate. At $1.0$ there is no decay. How do different decay rates influence the learning process?\n",
     "\n",
-    "-------"
+    "-------\n",
+    "\n",
+    "## **Challenge for motivated students**\n",
+    "The lecture introduces the concept of co-evolving mutation rates. Can you implement a function `co_evolving_mutation_rate` which implements this concept and integrate it in the evolutionary loop?"
    ]
   },
   {
@@ -232,13 +235,12 @@
     "# Evolution\n",
     "mean_vector = initialise_mean(num_parameters)\n",
     "\n",
-    "\n",
-    "# Control of mutation rate\n",
+    "# Exponential decaying learning rate\n",
+    "# works like a learning scheduler\n",
     "def exp_decay(sigma, sigma_decay, sigma_limit):\n",
     "    sigma = sigma * sigma_decay\n",
     "    return np.max((sigma_limit, sigma))\n",
     "\n",
-    "\n",
     "# Evolutionary loop\n",
     "for gen in range(num_generations):\n",
     "    population = generate_mutated_offspring(\n",