Code for reproducing most of the results in the paper:

Exploration in Online Advertising Systems with Deep Uncertainty-Aware Learning. Chao Du, Zhifeng Gao, Shuo Yuan, Lining Gao, Ziyan Li, Yifan Zeng, Xiaoqiang Zhu, Jian Xu, Kun Gai and Kuang-chih Lee. SIGKDD 2021.

This project is tested under the following environment settings:

• OS: Ubuntu 18.04.5 LTS
• CPU: Intel(R) Xeon(R) Platinum 8163
• GPU: N/A
• Python: 2.7.17
• tensorflow: 1.14.0
• numpy: 1.16.6

• The data can be downloaded from http://ml.cs.tsinghua.edu.cn/~chaodu/static/files/DUAL-data.tar.gz.
• After downloading, run tar -zxvf DUAL-data.tar.gz to extract the data files.
• The data folder should be placed in the base directory of the project.

We provide a convenient experiment launcher to produce results using multiple different seeds.

Try python batchrunner_dnn.py to train the DNN architecture with DUAL using seeds [0 - 31].

If everything goes well, this should reproduce the result "$0.7755 \pm 0.0020$" in Table 1.

Our code is developed based on the mouna99/dien project.