Isotropic Gaussian distribution
A probability distribution where all variables are independent and have equal variance in all dimensions; in Sigreg, this distribution is the target for representation learning because it maximizes entropy and ensures variables are maximally disentangled
As defined in 72Xj8k5WQX4
Other definitions
A probability distribution where variance is equal across all dimensions, visualized as a round ball in high-dimensional space; used in LeJepa to constrain embedding space geometry
What Is Yann LeCun Cooking? JEPA Explained Simply
Where it comes up · 2
- LeWorldModel: Stable End-to-End JEPA World Models From Pixels2 claims
- What Is Yann LeCun Cooking? JEPA Explained Simply1 claim
Most important claims · 3 of 3Search all claims for “Isotropic Gaussian distribution” →
- The core innovation in LayWM's regularization is SigReg (Sketched Isotropic Gaussian Regularizer), which enforces geometric diversity by forcing the AI's latent embeddings to match an isotropic Gaussian distribution—a perfectly diverse multi-dimensional bell curve—thereby mathematically forbidding representation collapse.LeWorldModel: Stable End-to-End JEPA World Models From Pixels
- The fundamental trade-off of forcing AI representations to match a rigid isotropic Gaussian distribution is that the mandatory geometric complexity causes LayWM to fail on simple tasks with low intrinsic dimensionality, such as a 2D dot navigating through a door (two-room environment).LeWorldModel: Stable End-to-End JEPA World Models From Pixels
- LeJepa constrains the overall geometry of the embedding space to follow an isotropic Gaussian distribution, meaning representation space spreads information evenly across dimensions so no direction collapses or dominates.What Is Yann LeCun Cooking? JEPA Explained Simply