CSDI<T>

Models & Types · AiDotNet.Finance.Forecasting.Foundation

CSDI — Conditional Score-based Diffusion Model for Probabilistic Time Series Imputation.

Inheritance: NeuralNetworkBase<T> → FinancialModelBase<T> → ForecastingModelBase<T> → TimeSeriesFoundationModelBase<T> → CSDI<T>

Implements: IGeneratedParameterRegistrar<T>

For Beginners

CSDI fills in missing data points in time series and forecasts future values using a diffusion process. Think of it like an artist restoring a damaged painting: it looks at the intact parts and intelligently fills in the gaps. Unlike simpler methods that fill one gap at a time, CSDI fills all missing values simultaneously, which produces more consistent and realistic results.

How It Works

CSDI uses score-based diffusion for non-autoregressive time series imputation and forecasting. It conditions on observed values using a transformer-based denoiser and generates all missing values simultaneously.

Reference: Tashiro et al., “CSDI: Conditional Score-based Diffusion Models for Probabilistic Time Series Imputation”, NeurIPS 2021.

Example

// Create a CSDI conditional score-based diffusion model for time series imputation
// Uses score-matching for probabilistic forecasting and missing value imputation
var architecture = new NeuralNetworkArchitecture<double>(
    inputType: InputType.OneDimensional,
    taskType: NeuralNetworkTaskType.Regression,
    inputHeight: 512, inputWidth: 1, inputDepth: 1, outputSize: 24);
            
// Training mode with conditional score-based diffusion
var model = new CSDI<double>(architecture);
            
// ONNX inference mode with pre-trained model
var onnxModel = new CSDI<double>(architecture, "csdi.onnx");

Constructors

Constructor Summary
CSDI(NeuralNetworkArchitecture<T> architecture, CSDIOptions<T> options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer, ILossFunction<T> lossFunction)
CSDI(NeuralNetworkArchitecture<T> architecture, string onnxModelPath, CSDIOptions<T> options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer, ILossFunction<T> lossFunction)

Properties

Property Summary
FoundationModelSize ModelSize { get; }
bool IsChannelIndependent { get; }
bool SupportsTraining { get; }
bool UseNativeMode { get; }
int MaxContextLength { get; }
int MaxPredictionHorizon { get; }
int NumFeatures { get; }
int PatchSize { get; }
int PredictionHorizon { get; }
int SequenceLength { get; }
int Stride { get; }

Methods

Method Summary
Dictionary<string, T> Evaluate(Tensor<T> predictions, Tensor<T> actuals)
Dictionary<string, T> GetFinancialMetrics()
ModelMetadata<T> GetModelMetadata()
Tensor<T> ApplyInstanceNormalization(Tensor<T> input)
Tensor<T> AutoregressiveForecast(Tensor<T> input, int steps)