TemporalGCNOptions<T>
Options & Configuration · AiDotNet.Models.Options
Configuration options for TemporalGCN (Temporal Graph Convolutional Network).
Inheritance: ModelOptions → RegressionOptions<T> → TimeSeriesRegressionOptions<T> → TemporalGCNOptions<T>
For Beginners
TemporalGCN captures two types of patterns simultaneously:
The Key Insight: Many real-world systems have both spatial structure (like a road network) and temporal dynamics (like traffic patterns over time). TemporalGCN learns both by alternating between spatial and temporal processing layers.
What is Temporal Graph Convolution?
- Standard GCN: Aggregates information from neighboring nodes at one time step
- TemporalGCN: Also captures how patterns evolve over time
- Combines spatial GCN layers with temporal recurrent (GRU/LSTM) layers
- Creates a “video” view of the graph, not just a “snapshot”
How TemporalGCN Works:
- Graph Convolution: Each node aggregates features from neighbors
- Temporal Recurrence: GRU/LSTM processes the sequence at each node
- Spatial-Temporal Stacking: Alternate spatial and temporal layers
- Prediction: Output future values at each node
TemporalGCN Architecture:
- GCN layers with Chebyshev polynomial approximation
- GRU cells for temporal modeling
- Batch normalization for stability
- Residual connections for gradient flow
Key Benefits:
- Jointly learns spatial and temporal patterns
- Handles dynamic graphs where edges change
- Scales to large graphs with sparse operations
- Works with irregular spatial structures
How It Works
TemporalGCN combines graph convolutional networks with recurrent neural networks to model both spatial and temporal dependencies in graph-structured time series data.
Reference: Zhao et al., “T-GCN: A Temporal Graph Convolutional Network for Traffic Prediction”, 2019.
https://arxiv.org/abs/1811.05320
Constructors
| Constructor | Summary |
|---|---|
TemporalGCNOptions() |
Initializes a new instance of the TemporalGCNOptions class with default values. |
TemporalGCNOptions(TemporalGCNOptions<T> other) |
Initializes a new instance by copying from another instance. |
Properties
| Property | Summary |
|---|---|
bool UseBatchNormalization { get; set; } |
Gets or sets whether to use batch normalization. |
bool UseResidualConnections { get; set; } |
Gets or sets whether to use residual (skip) connections. |
double DropoutRate { get; set; } |
Gets or sets the dropout rate for regularization. |
int ChebyshevOrder { get; set; } |
Gets or sets the Chebyshev polynomial order for graph convolution. |
int ForecastHorizon { get; set; } |
Gets or sets the forecast horizon. |
int HiddenDimension { get; set; } |
Gets or sets the hidden dimension for GCN and temporal layers. |
int NumGCNLayers { get; set; } |
Gets or sets the number of GCN layers. |
int NumNodes { get; set; } |
Gets or sets the number of nodes in the graph. |
int NumSamples { get; set; } |
Gets or sets the number of samples for uncertainty estimation. |
int NumTemporalLayers { get; set; } |
Gets or sets the number of temporal (recurrent) layers. |
int SequenceLength { get; set; } |
Gets or sets the sequence length (input time steps). |
string TemporalCellType { get; set; } |
Gets or sets the type of recurrent cell for temporal modeling. |