NTXentLoss<T>

Models & Types · AiDotNet.SelfSupervisedLearning.Losses

Normalized Temperature-scaled Cross Entropy Loss (NT-Xent) for contrastive learning.

Inheritance: ContrastiveLossBase<T> → NTXentLoss<T>

For Beginners

NT-Xent is the loss function used in SimCLR. It encourages representations of augmented views of the same image to be similar, while pushing apart representations of different images.

How It Works

How it works:

Loss formula:

Where τ (tau) is the temperature parameter.

Constructors

Constructor Summary
NTXentLoss(double temperature, bool normalize) Initializes a new instance of the NTXentLoss class.

Properties

Property Summary
double Temperature { get; } Gets the temperature parameter.

Methods

Method Summary
Tensor<T> ComputeLoss(Tensor<T> z1, Tensor<T> z2) Computes the NT-Xent loss for a batch of positive pairs.
ValueTuple<T, Tensor<T>, Tensor<T>> ComputeLossWithGradients(Tensor<T> z1, Tensor<T> z2) Computes the NT-Xent loss and returns gradients.