diff --git a/figures/a001.png b/figures/a001.png
new file mode 100644
index 0000000..ee8e48d
Binary files /dev/null and b/figures/a001.png differ
diff --git a/figures/a001_norm.png b/figures/a001_norm.png
new file mode 100644
index 0000000..43f24ae
Binary files /dev/null and b/figures/a001_norm.png differ
diff --git a/figures/a05.png b/figures/a05.png
new file mode 100644
index 0000000..d6fb7b9
Binary files /dev/null and b/figures/a05.png differ
diff --git a/figures/a05_norm.png b/figures/a05_norm.png
new file mode 100644
index 0000000..c7d44eb
Binary files /dev/null and b/figures/a05_norm.png differ
diff --git a/figures/a09.png b/figures/a09.png
new file mode 100644
index 0000000..a6f0348
Binary files /dev/null and b/figures/a09.png differ
diff --git a/figures/a09_norm.png b/figures/a09_norm.png
new file mode 100644
index 0000000..f953c82
Binary files /dev/null and b/figures/a09_norm.png differ
diff --git a/figures/pi1.png b/figures/pi1.png
new file mode 100644
index 0000000..cc6de92
Binary files /dev/null and b/figures/pi1.png differ
diff --git a/figures/vi1.png b/figures/vi1.png
new file mode 100644
index 0000000..9512c80
Binary files /dev/null and b/figures/vi1.png differ
diff --git a/main.py b/main.py
index f8c2838..ae447b8 100644
--- a/main.py
+++ b/main.py
@@ -162,7 +162,6 @@ def dynamic_programming(optimizer_step, g, terminator, return_history=False):
 
 
 def plot_j_policy_on_maze(j, policy, normalize=True):
-
     heatmap = np.full(MAZE.shape, np.nan, dtype=np.float64)
     if normalize:
         # Non-linear, but a discrete representation of different costs
@@ -221,7 +220,7 @@ if __name__ == '__main__':
             plt.figure(figsize=(9, 7))
             plt.subplots_adjust(top=0.9, bottom=0.05, left=0.1, right=0.95,
                                 wspace=0.1)
-            plt.suptitle('DISCOUNT: {}'.format(a) +
+            plt.suptitle('Discount: {}'.format(a) +
                          ('\nNormalized view' if normalize else ''))
             i = 1
             for opt in ['Value Iteration', 'Policy Iteration']:
@@ -242,24 +241,22 @@ if __name__ == '__main__':
 
     # Error graphs
     for opt in ['Value Iteration', 'Policy Iteration']:
-        plt.figure(figsize=(7, 10))
+        plt.figure(figsize=(6, 10))
         plt.figtext(0.5, 0.04, 'Number of iterations', ha='center',
                     fontsize='large')
-        plt.figtext(0.05, 0.5, 'Logarithm of cost RMSE', va='center',
+        plt.figtext(0.01, 0.5, 'Logarithm of cost RMSE', va='center',
                     rotation='vertical', fontsize='large')
-        plt.subplots_adjust(wspace=0.38, hspace=0.35, left=0.205, right=0.92,
+        plt.subplots_adjust(wspace=0.38, hspace=0.35, left=0.205, right=0.98,
                             top=0.9)
         plt.suptitle(opt)
         i = 1
         for a in [0.99, 0.7, 0.1]:
             for cost in ['g1', 'g2']:
-                # name = 'Cost: {}, discount: {}'.format(cost, a)
                 ALPHA = a
                 history = dynamic_programming(optimizers[opt], costs[cost],
                                               terminators[opt],
                                               return_history=True)
                 plt.subplot(3, 2, i)
-                # plt.gca().set_title(name)
                 plot_cost_history(history)
                 if i <= 2:
                     plt.gca().set_title('Cost: {}'.format(cost))
@@ -268,6 +265,5 @@ if __name__ == '__main__':
 
                 i += 1
 
-
     print('I ran in {} seconds'.format(time() - start))
     plt.show()
diff --git a/report.latex b/report.latex
index 6d5a507..c0c4b44 100644
--- a/report.latex
+++ b/report.latex
@@ -4,6 +4,8 @@
 \usepackage{fancyhdr}
 \pagestyle{fancy}
 \usepackage{lastpage}
+\usepackage{graphicx}
+% \graphicspath{{./figures}}
 \cfoot{Page \thepage\ of \pageref{LastPage}}
 \rhead{Pavel Lutskov, 03654990}
 \lhead{Programming Assignment}
@@ -46,18 +48,19 @@ be non-zero. Therefore, one would need a more space efficient representation of
 the transition probabilities, and therefore wouldn't be able to use a matrix
 library such as \textit{NumPy} for acceleration of computations.
 
-The one step costs in my implementation only depend on the target state,
-meaning $g(x, u, w) = g(f(x, u, w))$, therefore the cost functions are
+The one-step costs in my implementation only depend on the target state,
+meaning $g(x, u, w) = g(f(x, u, w))$, therefore the one-step cost functions are
 represented as vectors $G_x^1$ and $G_x^2$, where the goal state has a lower
 cost than the rest of the states, and the trap state incurs a high penalty.
 This formulation differs slightly from the formulation in the task, where for
 $g_2$ only the \textit{self-loop} in the final state is for free. However, this
 difference doesn't affect the resulting policy, and only has significant
-influence on the value function of the states directly adjacent to the goal
-state. If the cost did depend on the action taken to transit to the goal state
-(i.e.\ self-loop vs transition from the adjacent state), the cost couldn't have
-been stored as a vector, and instead a 2-D matrix would have been needed, which
-would have introduced unnecessary complexity to the code.
+influence on the cost function of the states directly adjacent to the goal
+state. If the one-step cost did depend on the action taken to transit to the
+goal state (i.e.\ self-loop vs transition from the adjacent state), the
+one-step cost couldn't have been stored as a vector, and instead a 2-D matrix
+would have been needed, which would have introduced unnecessary complexity to
+the code.
 
 A policy is implemented as a vector $\Pi_x$, where the $x$-th element of the
 vector contains the index of the action, that will be taken in state $x$.
@@ -71,36 +74,76 @@ discount factor $\alpha$, and the relation $\epsilon = \alpha^{|S|}$, where
 $|S|$ is the number of possible states, has been empirically found to provide
 good results.
 
-For visualization I used a non-linear scale for the value function. Each
-different value in the value vector was assigned a different color in order to
+\section{Algorithm inspection}
+
+For visualization I used a non-linear scale for the cost function. Each
+different value in the cost vector was assigned a different color in order to
 ensure, that for small values for $\alpha$ the distinct values could be clearly
 visible. The unnormalized representation is also provided as reference.
 
-\section{Algorithm inspection}
-
-If the termination criterion for Value Iteration is chosen correctly, i.e. the
+If the termination criterion for Value Iteration is chosen correctly, i.e.\ the
 algorithm only terminates when it converged to an optimal policy, then both PI
-and VI will result in the same policy. The cost $g_2$ is constantly shifted by
-$1$ relative to $g_1$, except for the trap state. For this reason $g_1$ and
-$g_2$ produce the same result for most $\alpha$, however the values of $\alpha$
-exist, for which the two costs produce different policies in the proximity of
-the trap. Generally, the behavior with both costs may differ, depending on the
-$\alpha$. For large $\alpha$ the algorithms may favor risking getting into the
-trap over going around it. For smaller $\alpha$ the resulting policy, on the
-contrary, is playing on the safe side.
+and VI will result in the same policy. The one-step cost $g_2$ is constantly
+shifted by $1$ relative to $g_1$, except for the trap state. For this reason
+$g_1$ and $g_2$ produce the same result for most $\alpha$, however the values
+of $\alpha$ exist, for which the two one-step costs produce different policies
+in the proximity of the trap. Generally, the behavior with both one-step costs
+may differ, depending on the $\alpha$. For large $\alpha$ the algorithms may
+favor risking getting into the trap over going around it. For smaller $\alpha$
+the resulting policy, on the contrary, is playing on the safe side.
 
 Furthermore, for very small $\alpha$, e.g.\ $0.01$, machine precision starts
 playing a role. The double precision floating point variable can store numbers
 of large range of magnitude, however the precision is limited by the 52-bit
-fractional part. The precision is not an issue for the cost $g_1$, because the
-negative cost of the goal state is propagated through the maze as a number of
-ever decreasing magnitude, since the one-step costs in the maze are $0$. For
-the cost $g_2$, however, the dominating term for the value function is the
-one-step cost of $1$ for the non-goal states, therefore the cost-free final
-state is propagated as an ever-decreasing additive term, and the distance of
-the propagation is restricted by the precision of the floating point variable
-used to store the value function. Hence, the algorithms may not converge to the
-optimal policy, when $g_2$ is used in conjunction with small values of
-$\alpha$.
+fractional part. The precision is not an issue for $g_1$, because the negative
+cost of the goal state is propagated through the maze as a number of ever
+decreasing magnitude, since the one-step costs in the maze are $0$. For $g_2$,
+however, the dominating term for the cost function is the one-step cost of $1$
+for the non-goal states, therefore the cost-free final state is propagated as
+an ever-decreasing additive term, and the distance of the propagation is
+restricted by the precision of the floating point variable used to store the
+cost function. Hence, the algorithms may not converge to the optimal policy,
+when $g_2$ is used in conjunction with small values of $\alpha$.
+
+For comparison of Value Iteration and Policy Iteration I used a wide range of
+$\alpha$, the values that I chose are $0.99$, $0.7$ and $0.1$. Using these
+values demonstrates the impact, that $\alpha$ has on the optimization. With
+large $\alpha$ it can be seen, that both algorithms stagnate for several
+iterations, after which they converge rapidly to the optimal policy and cost
+function. With decreasing $\alpha$ this effect becomes less pronounced, and the
+algorithms converge more steadily. From these graphs it is apparent, that
+Policy Iteration converges in two to three times less iterations than Value
+Iteration. Surprisingly, the number of iterations doesn't seem to depend on the
+discount factor, which could mean that the given maze problem is small and
+simple enough, so we don't have to care about choosing the $\alpha$ carefully.
+Furthermore, the one-step cost $g_2$ allows both algorithms to converge faster.
+
+It is natural, that PI converges in less iterations than VI, since policy is
+guaranteed to improve on each iteration. However, finding the exact cost
+function $J_{\pi_k}$ on each iteration can get expensive, when the state space
+grows. However, the given maze is small, so it is affordable to use the PI.
+
+\begin{figure}
+  \includegraphics[width=\linewidth]{figures/a09.png}
+  \includegraphics[width=\linewidth]{figures/a09_norm.png}
+\end{figure}
+
+\begin{figure}
+  \includegraphics[width=\linewidth]{figures/a05.png}
+  \includegraphics[width=\linewidth]{figures/a05_norm.png}
+\end{figure}
+
+\begin{figure}
+  \includegraphics[width=\linewidth]{figures/a001.png}
+  \includegraphics[width=\linewidth]{figures/a001_norm.png}
+\end{figure}
+
+\begin{figure}
+  \includegraphics[width=\linewidth]{figures/vi1.png}
+\end{figure}
+
+\begin{figure}
+  \includegraphics[width=\linewidth]{figures/pi1.png}
+\end{figure}
 
 \end{document}