This project is coded in Python 3.6 on Windows. Required packages include Numpy, Pandas and Matplotlib. Executes
pip3 install numpy matplotlib pandas
or
pip3 install -r requirements.txt
On Ubuntu 16.04, if getting error message "no module named _tkinter", then execute the following
sudo apt-get install python3-tk
Execute the following line to generate all three figures using new randomly generated training sets. Numerical results will also be displayed in the terminal.
python3 all_figures.py
Then generated figures will be in the folder figures folder. The file name of the reproduced figures will look like fig3_1550443283.png where 1550443283 is a time stamp, so that new generated figures will not overwrite old ones.
To generate each figure separately, run the following commands. Reproducing figure 3 may take half an hour to finish because it is trying different alphas.
python3 figure3.py
python3 figure4.py
python3 figure5.py
For reproducing Figure 4, the error TD(0) at α=0.6 could be between TD(0.3) and TD(0.8). Run the command python3 figure4.py three times or more to see a plot where the error TD(0) is above the error of TD(0.8), as shown in my report.
Execute the following lines to generate exactly the same figures in my report, using files in the data folder. (However, this might fail due to compatibility problem of different Numpy and Pickle versions.)
python3 all_figures.py -f
python3 figure3.py -f
python3 figure4.py -f
python3 figure5.py -f
figures: store geneated figure. Four figures used in the report are provided here. New figures generated by running the scripts are also stored here.
data: pickle files containing the experiment results that can generate the same figures in the report