Hi, I'm Subhransu.
This site collects short tutorials, explanations, and animations on topics I find interesting. Mostly robot path and motion planning, model predictive control, and learning-based approaches to robot motion such as vision-language-action models and world action models. In between, the math tips, tricks, and concepts that make those topics easier to understand.
- Importance samplingYou want an expectation under p, but you can only draw samples from q. Reweight each sample and the estimate stays unbiased. The catch is in the variance.