VisionTSOptions<T>

Options & Configuration · AiDotNet.Models.Options

Configuration options for VisionTS (Visual Masked Autoencoders as Zero-Shot Time Series Forecasters).

Inheritance: ModelOptions → RegressionOptions<T> → TimeSeriesRegressionOptions<T> → VisionTSOptions<T>

For Beginners

VisionTS brings image AI to time series:

Cross-Modal Transfer: VisionTS converts time series into 2D image-like representations, then uses a pretrained Visual MAE (originally trained on images) to process them. This leverages the massive pretraining of vision models for time series tasks.

How It Works:

  1. Convert time series to 2D patch grid (like an image)
  2. Mask some patches (MAE-style)
  3. Use the pretrained ViT encoder to process visible patches
  4. Decode masked patches to reconstruct/forecast the series

How It Works

VisionTS repurposes Visual Masked Autoencoders (MAE) for time series forecasting, demonstrating that vision foundation models can transfer effectively to the time series domain through cross-modal transfer.

Reference: “VisionTS: Visual Masked Autoencoders as Zero-Shot Time Series Forecasters”, ICML 2025.

Constructors

Constructor Summary
VisionTSOptions() Initializes a new instance with default values.
VisionTSOptions(VisionTSOptions<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.
double DropoutRate { get; set; } Gets or sets the dropout rate.
double LearningRate { get; set; } Gets or sets the Adam learning rate used by the paper’s one-epoch LayerNorm-only full-shot fine-tuning recipe.
double MaskRatio { get; set; } Gets or sets the mask ratio for MAE pretraining.
int ContextLength { get; set; } Gets or sets the context length.
int ForecastHorizon { get; set; } Gets or sets the forecast horizon.
int HiddenDimension { get; set; } Gets or sets the hidden dimension of the ViT encoder.
int ImageHeight { get; set; } Gets or sets the image height for 2D conversion.
int ImageWidth { get; set; } Gets or sets the image width for 2D conversion.
int IntermediateSize { get; set; } Gets or sets the intermediate size.
int NumHeads { get; set; } Gets or sets the number of attention heads.
int NumLayers { get; set; } Gets or sets the number of transformer layers.
int PatchLength { get; set; } Gets or sets the patch length for 2D patch grid conversion.