Hamiltonian Monte Carlo: probability as physics.
Endow particles with momentum, then let Hamilton’s equations (dq/dt = ∂H/∂p, dp/dt = −∂H/∂q) carve reversible, volume-preserving trajectories through phase space.
Diffusion (stochastic SDE sampler): erratic Brownian trajectories zigzagging through noise.
Flow Matching (deterministic ODE integrator): clean, straight-line paths to the data modes.
Same start, radically different dynamics.
The mixing time of exponential random graph models (ERGMs) measures how fast a Markov chain used to sample networks from an ERGM converges to its target distribution. ERGMs describe complex network data—social, biological, or information networks—by assigning probabilities to graphs based on features like edges, triangles, or degree patterns. In statistics, fast mixing is crucial because maximum likelihood and Bayesian inference for ERGMs rely on MCMC sampling; if the chain mixes slowly, estimates become biased and unreliable. Phase transitions in ERGMs can cause exponentially slow mixing, revealing when certain network structures are statistically unstable or unidentifiable. In machine learning and deep learning, ERGMs and their mixing behavior connect to training energy-based models and graph generative models, where poor mixing leads to mode collapse or bad samples. Understanding mixing time thus tells us when network learning, simulation, and prediction are computationally feasible in real-world graph data.
Image: https://t.co/JxlJHCotjT
Mathematical Mitosis. Luminous Buds of Instability. Almost reminds me of biological emergence.
"x>0 breaks the symmetry of the eigenvalues distribution."
Computation by Simone Conradi, @S_Conradi, Used with permission.
By the Koebe–Andreev–Thurston theorem, every finite triangulation of the plane is the tangent graph of some circle packing. This is a manim animation. It generates a finite graph with given number of boundary/interior points inside a bounding rectangle.
Our final module is out! We've covered the fundamentals of Information theory.
topics include -
> shannon entropy
> KL divergence
> information gain (decision trees)
> mutual information
> cross entropy loss
> GANs training
> Jensen-Shannon Divergence
> Perplexity