Results and Conclusions
Results
Fluxbox now effectively learns desired locations for windows and locations on the screen to avoid. It still places windows without overlap whenever possible. Due to the fast learning rate, it sometimes believes certain locations are more important than they are.Conclusions
The AI works, but its quality is not quite high enough yet to be released. Further adjustments to the learning rates and learning algorithm are required. Also, the weights are currently only adjusted when the window is moved or resized with the mouse. Keyboard manipulations will require more effort.
Once the code has reached a state where it can be released, there are many more options that could be included in future work. First, we could move and resize windows that are already open in order to make room for the new window. Also, Fluxbox supports tabbed windows, so two programs can be placed into the same window frame. Preferences for this behavior could and should be learned as well.