Zoomer

Implementation of the method proposed in the paper:

Zoomer: Improving and Accelerating Recommendation on Web-Scale Graphs via Regions of Interests.

Enviroment Requirement

The code has been tested running under Python 3.6.5. The requireed packages are as follows:

• XDL == 2.1
• Euler == 2.0

Configuration

To apply Zoomer to your own data, please:

• Configure model/xdl2_runner_zoomer_base_train.json.
• Set the hyperparameters of the model in model/zoomer_model_config.py.
• ROI-based Multi-Level Attention in this paper contains three major attention mechanisms: a flexible node-level feature embedding projection, an edge-level neighbor reweighing, and a semantic-level relation combination. You can enable or disable these attentions in model_option.

Data

• We used Alibaba's internal data and built our own industrial dataset in the experiment. Due to user privacy concerns and data security reasons, we can not disclose the data we use.
• You can apply Zoomer to your own datasets by following our input data's format.

Input Data Format

• User log data input

Field	Type	Comment
query_id	int64	ID of the query
user_id	int64	ID of the user
item_id	int64	ID of the item
labels	float32	value equals to 0 or 1, indicating whether user clicks item under the given query.

• Graph data

Field	Type	Comment
node_id	int64	node ID in the graph
node_type	string	type of the node in the graph, we use q(query), u(user) and i(item)
edge_type	string	type of the edge in the graph
feature_name	string	name for the feature of nodes, we use different features for different type of nodes

Run the Code

To try our code, you can input your own data, and directly run the following command:

python xdl2_runner_zoomer_base_train.py

References

• XDL
• Euler