Struct SimCLR 

pub struct SimCLR { /* private fields */ }

SimCLR framework (Chen et al., 2020).

Simple framework for contrastive learning of visual representations:

1. Apply two random augmentations to create positive pairs
2. Use NT-Xent loss (normalized temperature-scaled cross-entropy)

Reference

Chen, T., et al. (2020). A Simple Framework for Contrastive Learning of Visual Representations. ICML.

Implementations

impl SimCLR

pub fn new(temperature: f32, projection_dim: usize) -> SimCLR

Create SimCLR framework.

pub fn nt_xent_loss(&self, z_i: &[Vec<f32>], z_j: &[Vec<f32>]) -> f32

Compute NT-Xent loss for a batch of positive pairs.

Arguments

• z_i - First augmented view embeddings [batch_size, dim]
• z_j - Second augmented view embeddings [batch_size, dim]

For each sample i, its positive pair is z_j[i] and negatives are all other samples in the batch.

pub fn temperature(&self) -> f32

pub fn projection_dim(&self) -> usize

Trait Implementations

impl Clone for SimCLR

fn clone(&self) -> SimCLR

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for SimCLR

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.