Models Overview

This section provides detailed documentation for all supported models in PyGDA. Our framework offers a comprehensive collection of graph domain adaptation models, built on a flexible and extensible architecture.

Core Architecture

BaseGDA

The foundation of PyGDA's model architecture, providing:

  • Base class for all graph domain adaptation models
  • Core training and inference functionalities
  • Standardized interfaces for model customization
  • Common utility methods and configurations

Customization Guide

PyGDA is designed for easy customization and extension. To create your own model:

from pygda.models import BaseGDA

class CustomGDA(BaseGDA):
    def __init__(self, **kwargs):
        super().__init__(**kwargs)
        # Initialize your model components

    def fit(self, data):
        # Implement your training logic
        pass

    def predict(self, data):
        # Implement your inference logic
        return predictions

Key Features

  1. Flexible Base Architecture

    • Inherit from BaseGDA for consistent interface
    • Access to core functionalities and utilities
    • Standardized training and evaluation methods

  2. Easy Training Process

    • Use fit() method for model training
    • Support for custom hyperparameters
    • Flexible dataset input handling
    • Built-in optimization utilities

  3. Streamlined Evaluation

    • Simple predict() interface
    • Standardized performance metrics
    • Easy integration with evaluation pipelines

  4. Extensibility

    • Create custom model architectures
    • Add new training strategies
    • Implement domain-specific features
    • Integrate with existing PyGDA components

Usage Example

from pygda.models import A2GNN

# Initialize model
model = A2GNN(in_dim=100, hidden_dim=64, num_classes=7)

# Train model
model.fit(train_data)

# Make predictions
predictions = model.predict(test_data)

For detailed information about each model, please visit their respective documentation pages linked above.