Specialty
An arXiv review examines how linear systems can learn unknown parameters while operating, and why gathering information can carry a short-term performance cost.
Scientific publication3 min read
The paper follows controllability from the Kalman-family rank condition to bracket generation on finite-dimensional Lie groups. Examples in SO(3), SO(4) and SO(7) illustrate how local directions propagate, while the paper marks infinite-dimensional vector-field algebras as the point where the analogy breaks down.
Scientific publication4 min read
The preprint turns a worst-case safety question into an optimization over possible starting distributions. Its examples show that the resulting upper bounds depend on relaxation level, unsafe-region choice and moment uncertainty, and may remain loose when the model is approximated.
Scientific publication3 min read
A new arXiv preprint models how hard limits on terminal losses change portfolio exposure. Its simulations found selective reductions in risky positions, but the study does not test human investors or real markets.
Scientific publication3 min read
A preprint comparing direct and indirect stability certification reports that the indirect approach used fewer samples, ran substantially faster and remained informative when a simulated dataset included outliers. The evidence is theoretical and benchmark-based, so it does not establish universal superiority.
Scientific publication6 min read
A version 1 arXiv preprint develops a variational framework for studying how optimal values change as parameters move in a chosen direction. The results extend several tools for infinite-dimensional analysis but remain conditional on compactness, differentiability and regularity assumptions.
Scientific publication6 min read
A new arXiv preprint reports that an adaptive time-window method produced closely matching optimization results while reducing reported runtime and storage in the tested computational flow cases. The study also recovered cylinder-like wake structures and reduced pressure-drop objectives in U-bend cases, but did not test physical prototypes or three-dimensional turbulent flows.
Scientific publication4 min read