MGTSDOptions<T>
Options & Configuration · AiDotNet.Models.Options
Configuration options for MG-TSD (Multi-Granularity Time Series Diffusion).
Inheritance: ModelOptions → RegressionOptions<T> → TimeSeriesRegressionOptions<T> → MGTSDOptions<T>
For Beginners
MG-TSD improves on standard diffusion models by:
Multi-Granularity Guidance: Instead of denoising at a single resolution, MG-TSD processes the time series at multiple temporal granularities (e.g., hourly, daily, weekly). Coarser predictions capture long-range trends and guide finer predictions, resulting in more coherent forecasts across time scales.
Key Advantages:
- Better captures patterns at different time scales
- Coarse-to-fine guidance improves forecast coherence
- Produces calibrated probabilistic forecasts
How It Works
MG-TSD introduces a multi-granularity guidance diffusion model that captures temporal patterns at different scales. It uses a coarse-to-fine guidance mechanism where predictions at coarser granularities guide the fine-grained diffusion process.
Reference: Fan et al., “MG-TSD: Multi-Granularity Time Series Diffusion Models with Guided Learning Process”, ICLR 2024.
Constructors
| Constructor | Summary |
|---|---|
MGTSDOptions() |
Initializes a new instance with default values. |
MGTSDOptions(MGTSDOptions<T> other) |
Initializes a new instance by copying from another instance. |
Properties
| Property | Summary |
|---|---|
double BetaEnd { get; set; } |
Gets or sets the ending beta value for the noise schedule. |
double BetaStart { get; set; } |
Gets or sets the starting beta value for the noise schedule. |
double DropoutRate { get; set; } |
Gets or sets the dropout rate for regularization. |
double GuidanceWeight { get; set; } |
Gets or sets the weight for cross-granularity guidance. |
int ContextLength { get; set; } |
Gets or sets the number of historical time steps used as input context. |
int DiffusionSteps { get; set; } |
Gets or sets the number of diffusion (denoising) steps. |
int ForecastHorizon { get; set; } |
Gets or sets the number of future time steps to forecast. |
int HiddenDimension { get; set; } |
Gets or sets the hidden dimension of the transformer layers. |
int NumGranularities { get; set; } |
Gets or sets the number of temporal granularity levels for guidance. |
int NumHeads { get; set; } |
Gets or sets the number of attention heads. |
int NumLayers { get; set; } |
Gets or sets the number of transformer layers. |