odow / SDDP.jl

A JuMP extension for Stochastic Dual Dynamic Programming
https://sddp.dev
Other
304 stars 62 forks source link

Unclear tutorial: introductory theory #648

Closed sigmundholm closed 1 year ago

sigmundholm commented 1 year ago

In struggling with understanding the tutorial, specifically under the section Preliminaries: approximating the cost-to-go term. In the forth paragraph, could you specify in equations, the "reduced cost of the decision variable $\bar{x}$", which is referred to the "subgradient of the function $V_i$ with respect to $x$".

odow commented 1 year ago

Hi there. I don't know if I understand. What part is unclear?

odow commented 1 year ago

Would it help if I added the link [reduced cost](https://en.wikipedia.org/wiki/Reduced_cost)?

https://en.wikipedia.org/wiki/Reduced_cost

odow commented 1 year ago

Any suggestions? I'm very interested in improving the documentation for people learning SDDP, but it's hard for me to know what parts need work.

If not, I will give this a few days and then close.

sigmundholm commented 1 year ago

Yes, adding the link would help. It would also help stating in equations what the reduced cost is in this case.

sigmundholm commented 1 year ago

For me it would help if the text maybe was something like this: [...] then the reduced cost of the decision variable $\bar{x}$ is a subgradient of the function $V_i$ with respect to x (i.e. $\frac{d}{dx}V_i(x, .)$)!. Just to emphasize what a subgradient is.