Computational Physics seminars

Michael Borinsky introduces tropical quantum field theory as a simpler deformation limit that preserves useful information about an interacting theory. Its numerical algorithm computes massive-scalar perturbation theory with runtime polynomial in loop order and external-leg count, avoiding the factorial growth of a Feynman-diagram expansion and making very high loop orders feasible. Geometrically, amplitudes become volumes of graph-moduli spaces, whose recursions resemble those for Weil–Petersson volumes. The talk connects the limiting construction, geometry and algorithm.

Paco Villaescusa examines how simulations and artificial intelligence change the extraction of physical knowledge from precision cosmological surveys. The talk asks how much information lies in large-scale structure beyond power spectra, including nonlinear scales and higher-order statistics, and introduces the Quijote and Backlight simulation programmes. It then addresses the uncertain astrophysical processes that complicate small-scale inference through CAMELS, a framework combining cosmology, galaxy formation and machine learning. DREAMS is presented as a route to predicting dark-matter observables while accounting for baryonic uncertainties. The final part considers coupling simulations with increasingly capable AI: Denario is introduced as an agentic research system that generates hypotheses, designs analyses, writes and executes code, interprets results and refines research directions. The discussion considers what coordinated populations of agents and simulations spanning physical scales could make possible.

Rare molecular and physical transitions are difficult to observe in ordinary simulations. This talk casts estimation of the committor function as stochastic optimal control, using feedback to sample reactive trajectories. It develops backpropagation and off-policy Value Matching objectives, addresses trapping in intermediate metastable states, and compares committor, rate and equilibrium estimates on benchmark systems.

Benoit Assi examines limits on reliable theoretical predictions for present and future particle colliders. Low-order simulations of QCD radiation carry large uncertainties, hadronization is commonly modeled rather than derived, and effective theories contain more operators than measurements can constrain. Information theory and machine learning can incorporate improved calculations into simulations, quantify uncertainty, choose informative observables and advance hadronization theory. Geometry of effective-theory field space combines infinite operator families into finite physical quantities, while information measures identify the combinations experiments can resolve. The talk develops these complementary approaches to extracting the available information from LHC and future-collider data.

Maarten Stroeks examines ground-energy optimisation for fermionic Hamiltonians whose interactions are classical in a suitable basis, a setting motivated by electronic structure in chemistry and condensed matter. Despite QMA hardness, fermionic Gaussian states achieve an approximation ratio of at least one third independent of sparsity. Semidefinite algorithms handle several traceless and positive-semidefinite families, including fixed particle number. A construction that mixes covariance matrices, called a Gaussian blend, provides the technical foundation.

Biwei Dai explores generative models built around physical symmetries and multiscale structure. For cosmology, these models enable field-level likelihoods that extract more information from weak lensing than two-point statistics and expose model misspecification through generated samples. Related methods address lattice sampling and collider anomalies. Physically informed generative priors also support inverse problems, including intensity-interferometry reconstruction of black-hole accretion disks and analysis of pulsar-timing gravitational-wave backgrounds.