An analyst studies a common cold outbreak recorded on Tristan da Cunha in 1967. The table gives daily infected and recovered counts, but it does not reveal the transmission rate or the number initially susceptible. A simulator can generate plausible…
Bayesian GANs Learn to Sample Simulator Parameters Read More »
Before fitting a Bayesian regression network, a researcher draws functions from its prior. The intended behavior is smooth variation across the input range. The sampled curves instead look almost flat. The weight variances seemed reasonable. The functions they imply tell…
Finite Neural Networks Get Gaussian Error Bounds Read More »
A researcher increases the hidden width of a small classifier because the proof seems to demand it. Training already works with far fewer neurons. The gap raises an awkward question. Does the network need all those parameters, or does the…
Shallow ReLU Networks Get Sharper Generalization Bounds Read More »
Picture a thin film clamped in a tensile tester and stretched until it is visibly longer. About sixty displacement readings come off the sensors, and the question is where the film is stiff and where it is soft. Answering that…
LazyDINO Amortizes Bayesian Inversion for PDE Models Read More »
Picture a table with five columns and five hundred rows. The columns lean on each other in a chain, so column two says a lot about column three and very little about column five. You want a model of the…
Diffusion Models and Factorizable Density Estimation Read More »
A balancing robot, a building with a vibration damper and a chemical reactor all share one control problem. Pick the feedback gains that keep the system near its target at the lowest long run cost. Reinforcement learning attacks it by…
A Closed Form Hessian for Policy Gradient Control Read More »
A hyperspectral camera records hundreds of narrow colour bands for every pixel, and a trained analyst can read a crop or a rooftop out of that spectrum. Getting the analyst’s labels is the slow part. A model that must classify…
Label Propagation Carries the MIFOMO Results Read More »
The most important temperature in a gas turbine is one nobody can measure. The gas entering the first turbine stage is hotter than a thermocouple would survive, and the air flowing through the compressor has no meter either. Engineers estimate…
A Gas Turbine Model That Learns Only the Maps Read More »
Imagine a head and neck surgeon standing over a planning screen with a patient whose lower jaw must come out because of a tumor. Two virtual plans sit side by side. Both restore the outline of the face. Both use…
OsteoOpt Plans Jaw Reconstruction Around Bone Union Read More »
Think of a patient lying in a cardiac MRI scanner with an irregular heartbeat. The standard cine protocol asks for repeated breath holds and stitches the picture together from many beats, so an irregular rhythm smears the very thing the…
LAPANet Tracks Heart Motion Without Reconstructing MRI Read More »