LTP-MMF: Towards Long-term Provider Max-min Fairness Under Recommendation Feedback Loops
- Pricing Type: Free
- Price Range Start($):
The GitHub link is https://github.com/xuchen0427/ltp-mmf
This GitHub repository, “XuChen0427/LTP-MMF,” contains the implementation of LTP-MMF for SIGIR 2023. The code is intended for use in TOIS and provides the steam dataset, while other datasets can be downloaded from the URLs mentioned in the paper. The implementation is simulated with 256 users and 100 epochs due to space limitations. Users are advised to adjust parameters for different settings. To execute the code, run “python run_LTP-MMF.py.” The reported results include a final NDCG of 0.648, MMF of 0.502, and CTR of 0.555.
RFL means that recommender system can only receive feedback on exposed items from users and update recommender models incrementally based on this feedback.
The implementation of LTP-MMF in TOIS, no other use of the code is allowed! We here only provide steam dataset, other dataset please download them from the urls in the paepr Note that due to the limitation space of anonymous.4open.science, we only simulate it using 256 users and 100 epochs, please modify the parameters for other settings The result is: final NDCG:0.648 MMF:0.502 CTR:0.555
To balance efficiency and effectiveness, the vast majority of existing methods follow the two-pass approach, in which the first pass samples a fixed number of unobserved items by a simple static distribution and then the second pass selects the final negative items using a more sophisticated negative sampling strategy.