フィルターのクリア

Using fmincon with Matlab 2010 and 2016 returns different results

1 回表示 (過去 30 日間)
Maria445
Maria445 2016 年 11 月 21 日
回答済み: Walter Roberson 2016 年 11 月 21 日
I ran a code I used on a previous version of Matlab (2010) on the latest version (2016), but the results I got are different.
I need the code to find the parameters of a certain density function, using the maximum likelihood estimation. So I decided to use fmincon. The code I wrote is exactly the same on both Matlab versions, and so are the data I used. However, on the 2010 version I obtain a certain x vector and on the 2016 version I obtain a different one, even after having specified the same optimoptions in both versions.
Why is that so?

回答 (1 件)

Walter Roberson
Walter Roberson 2016 年 11 月 21 日
There are a number of different possibilities, including different default algorithms, differences in linear algebra libaries, and accumulation of round-off differences.
Does your function have some steep gradients? Exploration of those could be affected by round-off differences.

カテゴリ

Help Center および File ExchangeSolver Outputs and Iterative Display についてさらに検索

タグ

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by