Table of Contents

Interface IProbabilisticClassifier<T>

Namespace
AiDotNet.Interfaces
Assembly
AiDotNet.dll

Defines the interface for classifiers that can output probability estimates for each class.

public interface IProbabilisticClassifier<T> : IClassifier<T>, IFullModel<T, Matrix<T>, Vector<T>>, IModel<Matrix<T>, Vector<T>, ModelMetadata<T>>, IModelSerializer, ICheckpointableModel, IParameterizable<T, Matrix<T>, Vector<T>>, IFeatureAware, IFeatureImportance<T>, ICloneable<IFullModel<T, Matrix<T>, Vector<T>>>, IGradientComputable<T, Matrix<T>, Vector<T>>, IJitCompilable<T>

Type Parameters

T

The numeric type used for calculations (e.g., float, double).

Inherited Members
Extension Methods

Remarks

Probabilistic classifiers extend the basic classification interface by providing methods to obtain class probability estimates. This is useful for understanding the model's confidence in its predictions and for decision-making that considers uncertainty.

For Beginners: Some classifiers don't just say "this is category A" - they also tell you how confident they are.

For example, when classifying an email as spam:

  • A basic classifier might just say: "Spam"
  • A probabilistic classifier says: "90% spam, 10% not spam"

The probability information is valuable because:

  • You can see when the model is uncertain (50%/50% vs 99%/1%)
  • You can adjust the decision threshold (e.g., only mark as spam if >95% confident)
  • You can combine predictions from multiple models more effectively

Common probabilistic classifiers include:

  • Naive Bayes (naturally outputs probabilities)
  • Logistic Regression (outputs probabilities via sigmoid/softmax)
  • Random Forest (outputs probabilities via vote counting)

Methods

PredictLogProbabilities(Matrix<T>)

Predicts log-probabilities for each class.

Matrix<T> PredictLogProbabilities(Matrix<T> input)

Parameters

input Matrix<T>

The input features matrix where each row is a sample and each column is a feature.

Returns

Matrix<T>

A matrix where each row corresponds to an input sample and each column corresponds to a class. The values are the natural logarithm of the class probabilities.

Remarks

Log-probabilities are useful for numerical stability when working with very small probabilities, and for certain algorithms that work in log-space.

For Beginners: Log-probabilities are just probabilities transformed with the natural logarithm function.

Why use log-probabilities?

  1. Very small probabilities (like 0.0000001) become manageable numbers (like -16.1)
  2. Multiplying probabilities becomes adding log-probabilities (simpler and more stable)
  3. Many algorithms internally work in log-space for numerical stability

For example:

  • Probability 0.9 becomes log(0.9) = -0.105
  • Probability 0.1 becomes log(0.1) = -2.303
  • Probability 0.001 becomes log(0.001) = -6.908

Log-probabilities are always negative (since probabilities are between 0 and 1). Higher (less negative) values mean higher probability.

Unless you're doing advanced work, you probably want PredictProbabilities() instead.

PredictProbabilities(Matrix<T>)

Predicts class probabilities for each sample in the input.

Matrix<T> PredictProbabilities(Matrix<T> input)

Parameters

input Matrix<T>

The input features matrix where each row is a sample and each column is a feature.

Returns

Matrix<T>

A matrix where each row corresponds to an input sample and each column corresponds to a class. The values represent the probability of the sample belonging to each class. For each row, the probabilities sum to 1.0.

Remarks

This method computes the probability of each sample belonging to each possible class. The output matrix has shape [num_samples, num_classes], and each row sums to 1.0.

For Beginners: This method tells you how likely each sample is to belong to each category.

Example output for 3 samples with 3 classes (Cat, Dog, Bird):

Sample 1: [0.85, 0.10, 0.05]  // 85% Cat, 10% Dog, 5% Bird
Sample 2: [0.05, 0.90, 0.05]  // 5% Cat, 90% Dog, 5% Bird
Sample 3: [0.40, 0.35, 0.25]  // 40% Cat, 35% Dog, 25% Bird (uncertain!)

Notice that each row adds up to 1.0 (or 100%).

The third sample shows the model is uncertain - Cat is most likely, but it's not very confident. You might want to flag such cases for human review.

Edit this page