MOMENTOptions<T>
Options & Configuration · AiDotNet.Models.Options
Configuration options for MOMENT (Multi-task Optimization through Masked Encoding for Time series) foundation model.
Inheritance: ModelOptions → RegressionOptions<T> → TimeSeriesRegressionOptions<T> → MOMENTOptions<T>
For Beginners
MOMENT is like a Swiss Army knife for time series:
Key Innovation — Multi-Task Architecture: Unlike most time series models that only do forecasting, MOMENT handles 5 tasks:
- Forecasting: Predict future values
- Anomaly Detection: Find unusual patterns via reconstruction
- Classification: Label time series segments
- Imputation: Fill in missing values
- Embedding: Generate vector representations
How It Works:
- Divides input into patches (like Vision Transformer for images)
- Applies RevIN to handle different scales and distributions
- Uses a T5-style transformer encoder to process patches
- Task-specific heads generate outputs for each task type
Model Sizes:
- Small (~40M params): Fast experiments
- Base (~385M params): Strong general-purpose performance
- Large (~1B+ params): Maximum capacity
How It Works
MOMENT is a multi-task time series foundation model from Carnegie Mellon University. It uses a T5-based encoder-only transformer with patch embeddings and RevIN (Reversible Instance Normalization) to handle diverse time series tasks including forecasting, anomaly detection, classification, imputation, and embedding generation.
Reference: Goswami et al., “MOMENT: A Family of Open Time-Series Foundation Models”,
ICML 2024. https://arxiv.org/abs/2402.03885
Constructors
| Constructor | Summary |
|---|---|
MOMENTOptions() |
Initializes a new instance with default values. |
MOMENTOptions(MOMENTOptions<T> other) |
Initializes a new instance by copying from another instance. |
Properties
| Property | Summary |
|---|---|
FoundationModelSize ModelSize { get; set; } |
Gets or sets the model size variant. |
TimeSeriesFoundationModelTask Task { get; set; } |
Gets or sets the active task for the model. |
double DropoutRate { get; set; } |
Gets or sets the dropout rate for regularization. |
double MaskRatio { get; set; } |
Gets or sets the mask ratio for pretraining and imputation tasks. |
int ContextLength { get; set; } |
Gets or sets the context length (input sequence length). |
int ForecastHorizon { get; set; } |
Gets or sets the forecast horizon (prediction length). |
int HiddenDimension { get; set; } |
Gets or sets the hidden dimension of the T5 transformer. |
int IntermediateSize { get; set; } |
Gets or sets the intermediate size for the feed-forward network. |
int NumHeads { get; set; } |
Gets or sets the number of attention heads. |
int NumLayers { get; set; } |
Gets or sets the number of transformer encoder layers. |
int PatchLength { get; set; } |
Gets or sets the patch length for input tokenization. |
int? NumClasses { get; set; } |
Gets or sets the number of classification classes (for classification task only). |