This is the source code for the NIPS2018 paper: Bandit Learning with Implicit Feedback.
Version and Package requirement: python3 supported with numpy/scipy/matplotlib/pandas/coloredlogs. All package should be easily installed with conda.
Data: Please go to the following link and download the data named Xiaomu Questioning: http://moocdata.cn/data The description of the data is contained in the downlowded package.
Please unpack the package and move the data file(rawData.pkl) to the subdirectory ./data/ (configured in config.py) Then the following command should run the training phase:
python train.py
To see the figure of cumulative reward, try:
python plot.py
The figure should be plotted and saved.