the do-calculus does: computes interventional probabilities from observational data

log-probabilities are used instead of probabilities: avoids numerical underflow

Log-probabilities convert multiplications into additions, preventing numerical underflow

Manifold hypothesis

High-dimensional data lies on lower-dimensional manifolds

Lambda calculus

Lambda calculus represents data using only functions

Variational autoencoder

ELBO formula in variational inference

Tangent space

Tangent vectors map to manifold points via the exponential map

Causal model

Causal models use DAGs to represent causal relationships