[spelling mistake] chapeter -> chapter

chapter_reinforcement-learning/qlearning.md

@@ -67,7 +67,7 @@ It is important to note that when we pick an exploration that depends upon the c
 
 The dataset collected by the robot during Q-Learning grows with time. Both the exploration policy $\pi_e$ and the estimate $\hat{Q}$ evolve as the robot collects more data. This gives us a key insight into why Q-Learning works well. Consider a state $s$: if a particular action $a$ has a large value under the current estimate $\hat{Q}(s,a)$, then both the $\epsilon$-greedy and the softmax exploration policies have a larger probability of picking this action. If this action actually is *not* the ideal action, then the future states that arise from this action will have poor rewards. The next update of the Q-Learning objective will therefore reduce the value $\hat{Q}(s,a)$, which will reduce the probability of picking this action the next time the robot visits state $s$. Bad actions, e.g., ones whose value is overestimated in $\hat{Q}(s,a)$, are explored by the robot but their value is correct in the next update of the Q-Learning objective. Good actions, e.g., whose value $\hat{Q}(s, a)$ is large, are explored more often by the robot and thereby reinforced. This property can be used to show that Q-Learning can converge to the optimal policy even if it begins with a random policy $\pi_e$ :cite:`Watkins.Dayan.1992`.
 
-This ability to not only collect new data but also collect the right kind of data is the central feature of reinforcement learning algorithms, and this is what distinguishes them from supervised learning. Q-Learning, using deep neural networks (which we will see in the DQN chapeter later), is responsible for the resurgence of reinforcement learning :cite:`mnih2013playing`.
+This ability to not only collect new data but also collect the right kind of data is the central feature of reinforcement learning algorithms, and this is what distinguishes them from supervised learning. Q-Learning, using deep neural networks (which we will see in the DQN chapter later), is responsible for the resurgence of reinforcement learning :cite:`mnih2013playing`.
 
 ## Implementation of Q-Learning