You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
188 lines
7.5 KiB
188 lines
7.5 KiB
5 years ago
|
<!--TOC -->
|
||
|
|
|
||
|
|
- [Bayesian Graph Collaborative Filtering](#bayesian-graph-collaborative-filtering)
|
||
|
|
- [Model Architecture](#model-architecture)
|
||
|
|
- [Dataset](#dataset)
|
||
|
|
- [Features](#features)
|
||
|
|
- [Mixed Precision](#mixed-precision)
|
||
|
|
- [Environment Requirements](#environment-requirements)
|
||
|
|
- [Quick Start](#quick-start)
|
||
|
|
- [Script Description](#script-description)
|
||
|
|
- [Script and Sample Code](#script-and-sample-code)
|
||
|
|
- [Script Parameters](#script-parameters)
|
||
|
|
- [Training Process](#training-process)
|
||
|
|
- [Training](#training)
|
||
|
|
- [Model Description](#model-description)
|
||
|
|
- [Performance](#performance)
|
||
|
|
- [Description of random situation](#description-of-random-situation)
|
||
|
|
- [ModelZoo Homepage](#modelzoo-homepage)
|
||
|
|
<!--TOC -->
|
||
|
|
# [Bayesian Graph Collaborative Filtering](#contents)
|
||
|
|
|
||
|
|
Bayesian Graph Collaborative Filtering(BGCF) was proposed in 2020 by Sun J, Guo W, Zhang D et al. By naturally incorporating the
|
||
|
|
uncertainty in the user-item interaction graph shows excellent performance on Amazon recommendation dataset.This is an example of
|
||
|
|
training of BGCF with Amazon-Beauty dataset in MindSpore. More importantly, this is the first open source version for BGCF.
|
||
|
|
|
||
|
|
[Paper](https://dl.acm.org/doi/pdf/10.1145/3394486.3403254): Sun J, Guo W, Zhang D, et al. A Framework for Recommending Accurate and Diverse Items Using Bayesian Graph Convolutional Neural Networks[C]//Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 2020: 2030-2039.
|
||
|
|
|
||
|
|
# [Model Architecture](#contents)
|
||
|
|
|
||
|
|
Specially, BGCF contains two main modules. The first is sampling, which produce sample graphs based in node copying. Another module
|
||
|
|
aggregate the neighbors sampling from nodes consisting of mean aggregator and attention aggregator.
|
||
|
|
|
||
|
|
# [Dataset](#contents)
|
||
|
|
- Dataset size:
|
||
|
|
Statistics of dataset used are summarized as below:
|
||
|
|
|
||
|
|
| | Amazon-Beauty |
|
||
|
|
| ------------------ | -----------------------:|
|
||
|
|
| Task | Recommendation |
|
||
|
|
| # User | 7068 (1 graph) |
|
||
|
|
| # Item | 3570 |
|
||
|
|
| # Interaction | 79506 |
|
||
|
|
| # Training Data | 60818 |
|
||
|
|
| # Test Data | 18688 |
|
||
|
|
| # Density | 0.315% |
|
||
|
|
|
||
|
|
- Data Preparation
|
||
|
|
- Place the dataset to any path you want, the folder should include files as follows(we use Amazon-Beauty dataset as an example)"
|
||
|
|
```
|
||
|
|
.
|
||
|
|
└─data
|
||
|
|
├─ratings_Beauty.csv
|
||
|
|
```
|
||
|
|
|
||
|
|
- Generate dataset in mindrecord format for Amazon-Beauty.
|
||
|
|
```builddoutcfg
|
||
|
|
cd ./scripts
|
||
|
|
# SRC_PATH is the dataset file path you download.
|
||
|
|
sh run_process_data_ascend.sh [SRC_PATH]
|
||
|
|
```
|
||
|
|
|
||
|
|
- Launch
|
||
|
|
```
|
||
|
|
# Generate dataset in mindrecord format for Amazon-Beauty.
|
||
|
|
sh ./run_process_data_ascend.sh ./data
|
||
|
|
|
||
|
|
# [Features](#contents)
|
||
|
|
|
||
|
|
## Mixed Precision
|
||
|
|
|
||
|
|
To ultilize the strong computation power of Ascend chip, and accelerate the training process, the mixed training method is used. MindSpore is able to cope with FP32 inputs and FP16 operators. In BGCF example, the model is set to FP16 mode except for the loss calculation part.
|
||
|
|
|
||
|
|
# [Environment Requirements](#contents)
|
||
|
|
|
||
|
|
- Hardward (Ascend)
|
||
|
|
- Framework
|
||
|
|
- [MindSpore](https://www.mindspore.cn/install/en)
|
||
|
|
- For more information, please check the resources below:
|
||
|
|
- [MindSpore tutorials](https://www.mindspore.cn/tutorial/zh-CN/master/index.html)
|
||
|
|
- [MindSpore API](https://www.mindspore.cn/api/zh-CN/master/index.html)
|
||
|
|
|
||
|
|
# [Quick Start](#contents)
|
||
|
|
|
||
|
|
After installing MindSpore via the official website and Dataset is correctly generated, you can start training and evaluation as follows.
|
||
|
|
|
||
|
|
- running on Ascend
|
||
|
|
|
||
|
|
```
|
||
|
|
# run training example with Amazon-Beauty dataset
|
||
|
|
sh run_train_ascend.sh
|
||
|
|
```
|
||
|
|
|
||
|
|
# [Script Description](#contents)
|
||
|
|
|
||
|
|
## [Script and Sample Code](#contents)
|
||
|
|
|
||
|
|
```shell
|
||
|
|
.
|
||
|
|
└─bgcf
|
||
|
|
├─README.md
|
||
|
|
├─scripts
|
||
|
|
| ├─run_process_data_ascend.sh # Generate dataset in mindrecord format
|
||
|
|
| └─run_train_ascend.sh # Launch training
|
||
|
|
|
|
||
|
|
├─src
|
||
|
|
| ├─bgcf.py # BGCF model
|
||
|
|
| ├─callback.py # Callback function
|
||
|
|
| ├─config.py # Training configurations
|
||
|
|
| ├─dataset.py # Data preprocessing
|
||
|
|
| ├─metrics.py # Recommendation metrics
|
||
|
|
| └─utils.py # Utils for training bgcf
|
||
|
|
|
|
||
|
|
└─train.py # Train net
|
||
|
|
```
|
||
|
|
|
||
|
|
## [Script Parameters](#contents)
|
||
|
|
|
||
|
|
Parameters for both training and evaluation can be set in config.py.
|
||
|
|
|
||
|
|
- config for BGCF dataset
|
||
|
|
|
||
|
|
```python
|
||
|
|
"learning_rate": 0.001, # Learning rate
|
||
|
|
"num_epochs": 600, # Epoch sizes for training
|
||
|
|
"num_neg": 10, # Negative sampling rate
|
||
|
|
"raw_neighs": 40, # Num of sampling neighbors in raw graph
|
||
|
|
"gnew_neighs": 20, # Num of sampling neighbors in sample graph
|
||
|
|
"input_dim": 64, # User and item embedding dimension
|
||
|
|
"l2_coeff": 0.03 # l2 coefficient
|
||
|
|
"neighbor_dropout": [0.0, 0.2, 0.3]# Dropout ratio for different aggregation layer
|
||
|
|
"num_graphs":5 # Num of sample graph
|
||
|
|
```
|
||
|
|
|
||
|
|
## [Training Process](#contents)
|
||
|
|
|
||
|
|
### Training
|
||
|
|
|
||
|
|
- running on Ascend
|
||
|
|
```python
|
||
|
|
sh run_train_ascend.sh
|
||
|
|
```
|
||
|
|
|
||
|
|
Training result will be stored in the scripts path, whose folder name begins with "train". You can find the result like the
|
||
|
|
followings in log.
|
||
|
|
|
||
|
|
```python
|
||
|
|
Epoch 001 iter 12 loss 34696.242
|
||
|
|
Epoch 002 iter 12 loss 34275.508
|
||
|
|
Epoch 003 iter 12 loss 30620.635
|
||
|
|
Epoch 004 iter 12 loss 21628.908
|
||
|
|
|
||
|
|
...
|
||
|
|
Epoch 597 iter 12 loss 3662.3152
|
||
|
|
Epoch 598 iter 12 loss 3640.7612
|
||
|
|
Epoch 599 iter 12 loss 3654.9087
|
||
|
|
Epoch 600 iter 12 loss 3632.4585
|
||
|
|
epoch:600, recall_@10:0.10393, recall_@20:0.15669, ndcg_@10:0.07564, ndcg_@20:0.09343,
|
||
|
|
sedp_@10:0.01936, sedp_@20:0.01544, nov_@10:7.58599, nov_@20:7.79782
|
||
|
|
...
|
||
|
|
```
|
||
|
|
|
||
|
|
# [Model Description](#contents)
|
||
|
|
## [Performance](#contents)
|
||
|
|
|
||
|
|
| Parameter | BGCF |
|
||
|
|
| ------------------------------------ | ----------------------------------------- |
|
||
|
|
| Resource | Ascend 910 |
|
||
|
|
| uploaded Date | 09/04/2020(month/day/year) |
|
||
|
|
| MindSpore Version | 1.0 |
|
||
|
|
| Dataset | Amazon-Beauty |
|
||
|
|
| Training Parameter | epoch=600 |
|
||
|
|
| Optimizer | Adam |
|
||
|
|
| Loss Function | BPR loss |
|
||
|
|
| Recall@20 | 0.1534 |
|
||
|
|
| NDCG@20 | 0.0912 |
|
||
|
|
| Total time | 30min |
|
||
|
|
| Scripts | https://gitee.com/mindspore/mindspore/tree/master/model_zoo/official/gnn/bgcf |
|
||
|
|
|
||
|
|
# [Description of random situation](#contents)
|
||
|
|
|
||
|
|
BGCF model contains lots of dropout operations, if you want to disable dropout, set the neighbor_dropout to [0.0, 0.0, 0.0] in src/config.py.
|
||
|
|
|
||
|
|
# [ModelZoo Homepage](#contents)
|
||
|
|
|
||
|
|
Please check the official [homepage](http://gitee.com/mindspore/mindspore/tree/master/model_zoo).
|
||
|
|
|
||
|
|
|