DeepHitOptions<T>

Options & Configuration · AiDotNet.Models.Options

Configuration options for DeepHit survival analysis model.

For Beginners

While DeepSurv assumes that factors affect hazard rates proportionally (like “smoking doubles your risk at all times”), DeepHit makes no such assumption - it learns the actual probability of an event happening at each specific time point.

DeepHit is particularly useful when:

  • The proportional hazards assumption doesn’t hold
  • You have multiple competing risks (e.g., patient could die from disease OR treatment side effects)
  • You want to predict exact survival probabilities at specific time horizons
  • You need to handle complex, non-linear relationships in the data

For example: “What’s the probability this patient survives past 1 year? 2 years? 5 years?” DeepHit directly estimates these probabilities.

How It Works

DeepHit is a deep learning approach to survival analysis that directly learns the distribution of survival times without making assumptions like proportional hazards. It can also handle competing risks (multiple possible failure types).

Constructors

Constructor Summary
DeepHitOptions()

Properties

Property Summary
IActivationFunction<T> Activation { get; set; } Gets or sets the activation function for hidden layers.
bool UseBatchNormalization { get; set; } Gets or sets whether to use batch normalization.
double DropoutRate { get; set; } Gets or sets the dropout rate for regularization.
double L2Regularization { get; set; } Gets or sets the L2 regularization strength.
double LearningRate { get; set; } Gets or sets the learning rate for optimization.
double RankingSigma { get; set; } Gets or sets the sigma parameter for the ranking loss kernel.
double RankingWeight { get; set; } Gets or sets the weight for the ranking loss component.
double[] EvaluationHorizons { get; set; } Gets or sets the time horizons of interest for evaluation.
int BatchSize { get; set; } Gets or sets the batch size for training.
int Epochs { get; set; } Gets or sets the number of training epochs.
int HiddenLayerSize { get; set; } Gets or sets the number of neurons in each hidden layer.
int NumCauseLayers { get; set; } Gets or sets the number of hidden layers in each cause-specific sub-network.
int NumRisks { get; set; } Gets or sets the number of competing risks/causes.
int NumSharedLayers { get; set; } Gets or sets the number of hidden layers in the shared sub-network.
int NumTimeBins { get; set; } Gets or sets the number of discrete time bins for survival prediction.
int? EarlyStoppingPatience { get; set; } Gets or sets the patience for early stopping.
int? Seed { get; set; } Gets or sets the random seed for reproducibility.