Awesome

User Diverse Preferences Modeling By Multimodal Attentive Metric Learning

This is our implementation for the paper:

Fan Liu, Zhiyong Cheng*, Changchang, Yinglong Wang, Liqiang Nie*, Mohan Kankanhalli. User Diverse Preference Modeling via Multimodal Attentive Metric Learning. ACM International Conference on Multimedia (MM'19), Nice, France, 2019 (“*”= Corresponding author)

Please cite our ACMMM'19 paper if you use our codes. Thanks!

Environment Settings

• Tensorflow-gpu version: 1.3.0

Example to run the codes.

Run MAML.py

python MAML.py --dataset Office --num_neg 4 --eva_batches 400 --batch_size 5000 --hidden_layer_dim 256 --margin 1.6 --dropout 0.2 --feature_l2_reg 5.0

Dataset

We provide four processed datasets: Amazon-Office, Amazon-MenClothing, Amazon-WomenClothing, Amazon-Toys&Games.

train_csv:

• Train file.

test_csv:

• Test file.

asin_sample.json:

• Negative instances of items.

imge_feature.npy:

• Image features.

doc2vecFile:

• Text features.

All of the above files could be downloaded from :

• Google Drive Link

Last Update Date: AUG 15, 2021