This shows you the differences between two versions of the page.
chaosbook:relax [2011/01/04 20:37] predrag created "Relaxation for cyclists" chapter |
chaosbook:relax [2011/01/04 21:12] (current) predrag Jeff Moehlis talk at Northwestern Applied Math on control |
||
---|---|---|---|
Line 5: | Line 5: | ||
(ChaosBook.org blog, chapter [[http://chaosbook.org/paper.shtml#relax|Relaxation for cyclists]]) --- //[[predrag.cvitanovic@physics.gatech.edu|Predrag Cvitanovic]] 2011-01-04// | (ChaosBook.org blog, chapter [[http://chaosbook.org/paper.shtml#relax|Relaxation for cyclists]]) --- //[[predrag.cvitanovic@physics.gatech.edu|Predrag Cvitanovic]] 2011-01-04// | ||
- | ===== blah blah ===== | + | ===== Section 3: Least action method ===== |
- | ==== blah blah blah blah ==== | + | * **Predrag 2011-01-03** Heard Jeff Moehlis talk at Northwestern Applied Math on "neural" control. Learned this: |
+ | __Hamilton-Jacobi-Bellman PDEs__ minimize trajectory //time// under a range of time dependent external "control" inputs //u(t)//. This is a feed-forward, open lop control. Read Bardi, Capuzzo and Dolcetta 1997. | ||
+ | |||
+ | "arg min" means that the `value function' is minimized under a range of variations of the "argument" //u(t)//. | ||
+ | |||
+ | For Hodgkin-Huxley they minimize the power of electrical control input. That is done by __Euler-Lagrange equations__, with Lagrange multipliers of form //λ_1(t)(eqs of motion)//, //λ_2(t)(power constraint)//. They solve these by the shooting method. | ||
+ | |||
+ | ==== Subsection ==== | ||