Langevin dynamics does: adds noise to gradient descent to sample from a distribution

Langevin dynamics uses stochastic differential equations

Langevin dynamics incorporates randomness into the modeling of molecular systems, which helps simulate the effects of omitted degrees of freedom.

Example

In a simulation, Langevin dynamics can add random forces to the motion of particles, mimicking thermal fluctuations.

Remember this

This approach is crucial for understanding complex molecular behaviors that cannot be captured by deterministic models alone.

Text adapted from Wikipedia, licensed under CC BY-SA 4.0.

Related concepts

Diffusion model

q(x_t|x_{t-1}) adds Gaussian noise at each step

Stable Diffusion

Stable Diffusion generates images from text descriptions

Markov chain Monte Carlo

MCMC samples from complex posterior distributions

to standardize: when you need zero mean and unit variance for gradient-based optimization

Why do we need to make data uniform before training a model?

AdaGrad does: divides learning rate by sqrt of sum of squared gradients

How do we avoid overshooting in learning?

batch size affects generalization: larger batches find sharper minima

Larger batch sizes lead to sharper minima, enhancing generalization by providing more accurate gradient estimates

Swipe through 100 ML concepts daily