Latent Feature Dissection

Latent Feature Dissection

Latent Feature Dissection: Hypnotic Extraction of Disentangled Representation Subspaces

When we train deep neural networks, we often end up with rich latent representations that encode many factors of variation—but these factors are typic...

Latent Feature Dissection

The Dissociation Principle: Using Hypnotic Feature Separation to Resolve Confounded Latent Variables

When latent variables become confounded, models learn spurious correlations that break under distribution shift. The dissociation principle provides a...

Latent Feature Dissection

Hypnotic Decomposition: Isolating Latent Feature Signatures from Entangled Representation Manifolds

When we peer into the hidden layers of a deep neural network, we often encounter a dense, tangled web of activations. Features that we intuitively thi...