Score-Based Diffusion Models in Infinite Dimensions: A Malliavin Calculus Perspective

We develop a new mathematical framework for score-based diffusion modelling in infinite-dimensional separable Hilbert spaces, through Malliavin calculus, extending generative modelling techniques beyond the finite-dimensional setting. The forward diffusion process is formulated as a linear stochastic partial differential equation (SPDE) driven by space--time coloured noise with a trace-class covariance operator, ensuring well--posedness in arbitrary spatial dimensions. Building on Malliavin calculus and an infinite-dimensional extension of the Bismut--Elworthy--Li formula, we derive an exact, closed-form expression for the score function (the Fréchet derivative of the log-density) without relying on finite-dimensional projections or discretisation. Our operator--theoretic approach preserves the intrinsic geometry of Hilbert spaces and accommodates general trace-class operators, thereby incorporating spatially correlated noise without assuming semigroup invertibility. The framework also connects score estimation to functional data analysis, enabling computational strategies such as kernel methods and neural operator architectures.