Light

dongmanqing / code-for-mamo Goto Github PK

View Code? Open in Web Editor NEW

53.0 2.0 14.0 1.32 MB

The code for paper MAMO: Memory-Augmented Meta-Optimization for Cold-start Recommendation

License: Apache License 2.0

Python 100.00%

code-for-mamo's Introduction

Code-for-MAMO

Code for paper MAMO: Memory-Augmented Meta-Optimization for Cold-start Recommendation.

Requirements

python 3.6+

Packages

pytorch
numpy
pandas

Dataset

The raw datasets could be downloaded from:

Put the dataset into the according data_raw folder.
Create a folder named data_processed, you can process the raw datasets via python3 prepareDataset.py

Here we only give the processing code for the MovieLens dataset, please write your own code for processing Bookcrossing dataset with the similar functions presented in prepareMovielens.py

The structure of the processed dataset:

- data_processed

  - bookcrossing
    - raw
      sample_1_x1.p
      sample_1_x2.p
      sample_1_y.p
      sample_1_y0.p
      ...
    item_dict.p
    item_state_ids.p
    ratings_sorted.p
    user_dict.p
    user_state_ids.p
   
  - movielens
    - raw
      sample_1_x1.p
      sample_1_x2.p
      sample_1_y.p
      sample_1_y0.p
      ...
    item_dict.p
    item_state_ids.p
    ratings_sorted.p
    user_dict.p
    user_state_ids.p

Model training

The structure of our code:

- prepare_data
  prepareBookcrossing.py
  prepareList.py
  prepareMovielens.py
- modules
  info_embedding.py
  input_loading.py
  memories.py
  rec_model.py
configs.py
mamoRec.py
models.py
prepareDataset.py
utils.py

Run the codes in mamoRec.py for training the model:

if __name__ == '__main__':
    MAMRec('movielens')

Some tips

This version of code runs over all training or testing users, which may take about half an hour for one epoch on a Linux server with NVIDIA TITAN X. So you can revise the code for updating the parameters via batches of users and using parallel computing.
You can del the used variables to save the computation cost. If you have any suggestions on saving the computation cost, I'm happy to receive your emails.

Citation

If you use this code, please consider to cite the following paper:

@inproceedings{dong2020mamo,
  title={MAMO: Memory-Augmented Meta-Optimization for Cold-start Recommendation},
  author={Manqing, Dong and Feng, Yuan and Lina, Yao and Xiwei, Xu and Liming, Zhu},
  booktitle={26th SIGKDD Conference on Knowledge Discovery and Data Mining},
  year={2020}
}

code-for-mamo's People

Contributors

Stargazers

Watchers

Forkers

huoxusg helenligit gry123456 xu94-nlp hongtaicao leerinji juyongjiang pooyamoini mingyunhu colbyzhou aliang-rec luckydl21 crystal22 markhsia

code-for-mamo's Issues

About the data processing process to get 4 sub-datasets

Hi, I'm reading your paper MAMO and try to follow the experiment.
However, I have a question about the data processing process. In the function data_generation() in prepareDataset.py, it seems that all items of one user are mixed, whether it is warm or cold. Is further processing required to get 4 sub-datasets? And the training process doesn't seem to differentiate between the four statuses. If could, can you show the further data processing process and update the code?

Looking forward to your reply.

the document loss

why i can not found the prepareBookcrossing.py,please help me

Question about Video memory usage

i have read the paper of MAMO. it's a great work.Video memory usage increases with the increase of tasks, even if I delete some useless variables。After my debugging, I found that modifying the 《user_mem_init》 function in the model module can avoid this problem。

with torch.no_grad():
   pu = loading_model(u_x1)
   personalized_bias_term,` att_values = feature_mem.read_head(pu, alpha)

Question about the experiments result.

hello,
i have read the paper of MAMO. it's a great work. but i still have some question about it.
in this paper, you said “We can see that our method shows stable and well performance in different scenarios, which shows the effectiveness of the meta-augmented meta-optimization strategy: learning a personalized initialization of local parameters from the memories.”, but in the Table 1, i found the result of MAMO is not always better than other baseline.
thanks for your reading and looking forward to your reply.

Are the 4 scenarios for only testing or training?

Hi, I have read the paper MAMO and found it an interesting and novel paper.
Now I'm trying to reproduce it but not sure whether the 4 scenarios, including WW, WC, CW, CC, are used for training or not. If for training, the number of training samples will be too small for CW and CC scenarios. If only for testing, does that mean training on mixed setting and testing on separate settings?

Looking forward to your reply.

run problems

where can i find the file? thank you

support for multi attribute user features

Hi,
Like movie can have multiple genre and they are supported by the model, however i didnt find reference in code where the model provides support for such features(multi attribute) for users also. Can you pls elaborate if model support such user features?

Recommend Projects

React

A declarative, efficient, and flexible JavaScript library for building user interfaces.
Vue.js

🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript

TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
TensorFlow

An Open Source Machine Learning Framework for Everyone
Django

The Web framework for perfectionists with deadlines.
Laravel

A PHP framework for web artisans
D3

Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

javascript

JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
web

Some thing interesting about web. New door for the world.
server

A server is a program made to process requests and deliver data to clients.
Machine learning

Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Visualization

Some thing interesting about visualization, use data art
Game

Some thing interesting about game, make everyone happy.

Recommend Org

Facebook

We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft

Open source projects and samples from Microsoft.
Google

Google ❤️ Open Source for everyone.
Alibaba

Alibaba Open Source for everyone
D3

Data-Driven Documents codes.
Tencent

China tencent open source team.