Neelay Ranjan · generative-modeling researcher

energy-based models · diffusion · flight-path generation

Illustrative Langevin simulation. The name is the ground state of an energy landscape carved from its own letterforms: particles descend into the wells and settle. Impulses repel them out; they re-anneal back in. Drag to pick up a cluster. Hand-built simulation, not a trained model.

NASA Ames · Regenstrief Institute

These are real models I trained myself, running live in your browser. The first demo is the exception: a model this size would mean a 50 MB download the moment you arrive, before anything moved, so it replays real frames I exported instead.

·.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.··.:-=+=:.·

continuous-diffusion --digit 7booting

sdedit --digit 7 --strength 0.6booting

entropy-chess --engine ebm --search argminbooting

jepabooting

sample-space --compare ddpm,flowbooting

DDPM vs. flow matching. Both learn to turn noise into data, but they take different routes there. A diffusion model (DDPM) reverses a stochastic noising process: sampling is a random walk that removes a little noise at each of many steps, so the path from noise to sample is jagged and takes a different route every run. Flow matching instead learns a velocity field and follows it as a deterministic ODE. The trajectory is smooth, repeatable, and much straighter, which is why it can sample in far fewer steps. Straighten the paths further (rectified flow / reflow) and they approach straight line segments, collapsing dozens of steps into a handful.

The panels above are illustrative: hand-drawn fields on a 2D toy distribution, not a trained model. In two dimensions these trajectories can be made genuinely real; in the high-dimensional space of actual image models the same picture becomes a projection.