Better scaling for reward plot + axis labels #216
Conversation
|
Looks good. Which would you prefer, the actual reward, or the moving average? Could you possible add an argument to |
|
Yes, that's what I was thinking. Would it be ok to use a |
|
Yeah, that's fine. We may want to remove the pandas dependency in the future, but for now it's okay. |
| if self.reward_im is None and self.reward_ax is None: | ||
| self.reward_im, self.reward_ax = plt.subplots() | ||
| self.reward_ax.set_title('Accumulated reward') | ||
| self.reward_ax.set_xlabel('Episode') | ||
| self.reward_ax.set_ylabel('Reward') | ||
| self.reward_plot, = self.reward_ax.plot(self.reward_list) | ||
| self.reward_plot, = self.reward_ax.plot(reward_list_) |
There was a problem hiding this comment.
If self.reward_window is None, is reward_list_ created?
There was a problem hiding this comment.
Yes, that is done in the above if-else statement. If self.reward_window is None, reward_list_ = self.reward_list[:].
There was a problem hiding this comment.
Nice, I missed that for some reason.
| if self.reward_im is None and self.reward_ax is None: | ||
| self.reward_im, self.reward_ax = plt.subplots() | ||
| self.reward_ax.set_title('Accumulated reward') | ||
| self.reward_ax.set_xlabel('Episode') | ||
| self.reward_ax.set_ylabel('Reward') | ||
| self.reward_plot, = self.reward_ax.plot(self.reward_list) | ||
| self.reward_plot, = self.reward_ax.plot(reward_list_) |
There was a problem hiding this comment.
Nice, I missed that for some reason.
In case accumulated rewards are constant for some episodes, the horizontal line is obscured since it's right at the figures edge when using
minandmaxforset_ylim. Instead, we could do automatic scaling withrelimandautoscale_view, which works well and leaves margins. See for example https://stackoverflow.com/questions/10984085/automatically-rescale-ylim-and-xlim-in-matplotlib. If interested, I could also add the option to plot a moving average of reward instead of the actual rewards, since those tend to be very shaky.