Gaurav Khanal
Machine/deep learning, information geometry, topology, and optimization. I am interested in all things geometry behind probability, uncertainty, and learning systems.
Current writing
I write technical essays on mathematics, machine learning, geometry, and related ideas. My current long-form series is The Natural Geometry of Learning, beginning with Fisher information, optimal transport, KL geometry, and Schrödinger bridges.
The Natural Geometry of Learning
A long-form essay series on information geometry, optimal transport, KL divergence, Schrödinger bridges, and learning.
Research notes
Technical notes, paper summaries, derivations, and working ideas in mathematics and machine learning.
Projects
Selected work in machine learning, mathematical modeling, computation, and visualization.
Recent focus
My current interests sit at the intersection of geometry, probability, and learning:
- information geometry and Fisher metrics
- optimal transport and Wasserstein geometry
- KL divergence, Bregman geometry, and exponential families
- Schrödinger bridges and Langevin dynamics
- geometric viewpoints on machine/deep learning
Start here
If you are here for the writing project, start with the blog series: