Gradient systems associated with probability distributions
- 50 Downloads
Gradient systems on manifolds of various probability distributions are presented. It is shown that the gradient systems can be linearized by the Legendre transformation. It follows that the corresponding flows on the manifolds converge to equilibrium points of potential functions exponentially. It is proved that gradient systems on manifolds of even dimensions are completely integrable Hamiltonian systems. Especially, the gradient system for the Gaussian distribution admits a Lax pair representation.
Key wordsmanifolds of probability distributions gradient systems Hamiltonian systems Lax pair representation linear and nonlinear programming problems
Unable to display preview. Download preview PDF.
- A. Fujiwara, Dynamical systems on statistical models. Recent Developments and Perspectives in Nonlinear Dynamical Systems (eds. Y. Nakamura, K. Takasaki and K. Nagatomo), RIMS Kokyuroku No. 822, Kyoto Univ., Kyoto 1993, 32–42.Google Scholar
- N. Karmarkar, Riemannian geometry underlying interior-point methods for linear programming. Mathematical Developments Arising from Linear Programming (eds. J.C. Lagarias and M.J. Todd), Contemp. Math., Vol. 114, Amer. Math. Soc., Providence, 1990, 51–75.Google Scholar
- Y. Nakamura, Lax equations associated with a least squares problem and compact Lie algebras. Recent Developments in Differential Geometry (ed. K. Shiohama), Adv. Stud. Pure Math., Vol. 22, Math. Soc. Japan, 1993, 213–229.Google Scholar
- K. Tanabe and T. Tsuchiya, New geometry in linear programming (in Japanese). Math. Sci., No. 303, 1988, 32–37.Google Scholar