Summer School Course: Introduction to Learning-to-Rank

1. Questionnaire

In order to make the course interesting and easy-to-follow, it would be highly appreciated if you could provide some details on the following questions, thanks for your collaboration and time!

Before the course

https://wj.qq.com/s2/8749381/75d6/
After the course

https://wj.qq.com/s2/8749791/46a5/

2. Expected preliminary preparations

If you're not familiar with python, it will be quite helpful to learn some basics on python and numpy before the course based on the resources on the web, especially getting the environment for testing python programs prepared, since we plan to share some example python programs as a reference.

3. Required exercises after each lecture (dynamically updated)

Exercises after Lecture-1

https://wj.qq.com/s2/8749683/ba0d/
Exercises after Lecture-2

https://wj.qq.com/s2/8751303/c5d0/
Exercises after Lecture-4

https://wj.qq.com/s2/8783980/4285/

4. Challenging topics (optional)

The following topics can be finished by a number of students as a team.

Topic-1: Re-implement the newly proposed method in SIGIR-2021 using PyTorch.
- author: Oosterhuis, Harrie.
- title: Computationally Efficient Optimization of Plackett-Luce Ranking Models for Relevance and Fairness.
- conference: proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval, 1023–1032, 2021.
Note: this paper won the best paper award.
Topic-2: Try to extend the aforementioned method in topic-1.
Topic-3: Try to improve the following method in SIGIR-2021.
- author: Qu, Leigang and Liu, Meng and Wu, Jianlong and Gao, Zan and Nie, Liqiang.
- title: Dynamic Modality Interaction Modeling for Image-Text Retrieval.
- conference: proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval, 1104–1113, 2021.
Note: this paper by researchers from Shandong University won the best student paper award.
Topic-4: Four-digit MNIST sorting
- Try to explore how to perform four-digit MNIST sorting based on an open-sourced method or explore a new method.
- reference paper:
- Section 5.1 in Differentiable Sorting Networks for Scalable Sorting and Ranking Supervision
- links for the paper: (1) http://proceedings.mlr.press/v139/petersen21a/petersen21a.pdf (2) https://arxiv.org/abs/2105.04019

5 QA

If you have any questions, please do not hesitate to contact via the following ways:

A short message through the WeChat group;
Directly pose an issue on this Github project;

6 Reference slides and programs

Some necessary programs and slides will be shared via this Github project.

Noteworthy, it is prohibited to distribute at will without permission.

ii-research-yu/LTR_SummerSchool