quasi-independence

Markov chain mixing. SHould I run seperate paths or just wait long and kill data?

I have an experiment can run it N times but its expensive. Instead run it once and wait.

Simchowitz et al. (n.d.)

References

Simchowitz, Max, Horia Mania, Stephen Tu, Michael I. Jordan, and Benjamin Recht. n.d. “Learning Without Mixing: Towards a Sharp Analysis of Linear System Identification.” https://doi.org/10.48550/arXiv.1802.08334.