赵丰 (Zhao Feng)
赵丰 (Zhao Feng)
Need it
`scipy.integrate.solve_ivp` makes this feature compulsory for `DP5`,`BS3` and `DP853`. I made [a PR](https://github.com/scipy/scipy/pull/14500) on this issue. By the way, I could not get exactly the same solution result for `BS3`...
I found that the accuracy defects comes from `OrdinaryDiffEq`. The critical line is [here](https://github.com/SciML/OrdinaryDiffEq.jl/blob/e3da7e00415dc55730e8e3c35438b39b13a08378/src/integrators/controllers.jl#L65). In the `IController`, the power operation is computed using float32 number. Though it could speed up...
> > I found that the accuracy defects > > I can obtain the almost the same numerical results with scipy.integrate.solve_ivp for BS3 and DP5. > > That's not a...
> We could add an option to swap back the floating point pow implementation in the controllers, though using a same-accuracy default which is faster is fine. IIRC, it's a...
Same problem persists in Ubuntu 18.04.
看了你的博客非技术的文章，受益匪淺，感谢分享。或许考虑把技术与非技术用某种方式（比如打标签）分开？ 对于不是专门做前端的同学很多技术方面的文章可能帮助不大。 最近网络舆论上好像有一些八卦，希望你能早日渡过难关。
So it means that current implementation does not give s-t cut information?
*.proxy.ustclug.org is planned to be closed?