search

eyounx / VirtualTaobao

Virtual-Taobao simulators with OpenAI Gym interface
520 stars 98 forks source link

how to use it on personal dataset in my search Engine #3

Closed jjw-AI closed 4 years ago

jjw-AI commented 4 years ago

I am building an demo e-commerce search Engine for recommending item list to users with DDPG reinforcement learning algorithm predicting . But I am confused with the virtualTaobao as a virtual training environment to help me apply FL to the online system. 1 what The virtualTao help me for the above goal? 2 how to use the VirturalTao adapt my dataset for giving score to get topK list,can you give a example step for me ? 3 [virtualTB/SupervisedLearning/dataset.txt ] can you give a detail explanation for the features , label ,and the number of clicks. AND why the lables is multi length？

eyounx commented 4 years ago

This Virtual-Taobao project provides an environment for testing recommendation algorithms. In other words, this is a (live) benchmark.

If you want to apply reinforcement learning to your own engine, you may want to follow the Virtual-Taobao paper to build a simulator for your engine, and then train your RL agent in that simulator.

jjw-AI commented 4 years ago

Thanks!........ It seems that I need build a simulator for my dataset. Can you give a detail info/explanation about how to define the state and action in the ReadMe? There are so many dimensions , What do they mean? ![image](https://user-images.githubusercontent.com/11384622/71078261-f6bad480-21c3-11ea-9354-be880cbd37dc.png)