https://github.com/xidongbo/AITM

TensorFlow implementation of Adaptive Information Transfer Multi-task (AITM) framework. Code for the paper accepted by KDD21: Modeling the Sequential Dependence among Audience Multi-step Conversions with Multi-task Learning in Targeted Display Advertising. [https://arxiv.org/abs/2105.08489]
https://github.com/xidongbo/AITM

Last synced: 3 months ago
JSON representation

TensorFlow implementation of Adaptive Information Transfer Multi-task (AITM) framework. Code for the paper accepted by KDD21: Modeling the Sequential Dependence among Audience Multi-step Conversions with Multi-task Learning in Targeted Display Advertising. [https://arxiv.org/abs/2105.08489]

• Host: GitHub
• URL: https://github.com/xidongbo/AITM
• Owner: xidongbo
• License: mit
• Created: 2021-01-19T02:44:00.000Z (almost 4 years ago)
• Default Branch: main
• Last Pushed: 2023-10-19T11:49:11.000Z (about 1 year ago)
• Last Synced: 2024-06-24T05:53:55.190Z (5 months ago)
• Language: Python
• Homepage:
• Size: 71.3 KB
• Stars: 103
• Watchers: 1
• Forks: 30
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• StarryDivineSky - xidongbo/AITM - CCP阿里巴巴点击和转化预测数据集](https://tianchi.aliyun.com/datalab/dataSet.html?dataId=408)。[pytorch实现](https://github.com/adtalos/AITM-torch) (其他_推荐系统 / 网络服务_其他)

README

        
# AITM

TensorFlow implementation of Adaptive Information Transfer Multi-task (AITM) framework.  

Code for the paper accepted by KDD21: 

Modeling the Sequential Dependence among Audience Multi-step Conversions with Multi-task Learning in Targeted Display Advertising.

[https://arxiv.org/abs/2105.08489]

# Note(New)

Recently we reformatted the model code as ```AITM_standard.py```. If you want to run the model on multiple tasks (more than two), you can directly pass in the parameter ```--num_tasks``` in ```AITM_standard.py```. But you need to configure the ```config.csv``` file to specify the size of the feature dictionary.

# Requirement

python==3.6  

tensorflow-gpu==1.10.1  

# Dataset

We use the public Ali-CCP (Alibaba Click and Conversion Prediction) dataset. [https://tianchi.aliyun.com/datalab/dataSet.html?dataId=408].

Please download and unzip the dataset first.

Split the data to train/validation/test files to run the codes directly:

```

python process_public_dataset.py

```

# Example to run the model

```

python AITM.py --embedding_dim 5 --lr 1e-3 --early_stop 1 --lamda 1e-6 --prefix AITM --weight 0.6

```

The instruction of commands has been clearly stated in the codes (see the parse_args function).

# Reference

If you are interested in the code, please cite our paper:

```

Xi D, Chen Z, Yan P, et al. Modeling the sequential dependence among audience multi-step conversions with multi-task learning in targeted display advertising[C]//Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining. 2021: 3745-3755.

```

or in bibtex style:

```

@inproceedings{xi2021modeling,

  title={Modeling the sequential dependence among audience multi-step conversions with multi-task learning in targeted display advertising},

  author={Xi, Dongbo and Chen, Zhen and Yan, Peng and Zhang, Yinger and Zhu, Yongchun and Zhuang, Fuzhen and Chen, Yu},

  booktitle={Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery \& Data Mining},

  pages={3745--3755},

  year={2021}

}

```

# Other unofficial implementations for reference:

## A PyTorch implementation of multi-task recommendation models

[https://github.com/easezyc/Multitask-Recommendation-Library]

Last Update Date: Oct. 19, 2023 (UTC+8)