sparse matrix with higher than double precision

Question

Jacob Bresler 2016-8-1

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/297919-sparse-matrix-with-higher-than-double-precision

编辑： Thorsten 2016-8-1

hey guys, so I am currently working on a problem where all of my matrices are banded, but I am using the symbolic toolbox to set my floating point precision higher than that of double (I kind of have to since some of the numbers I am dealing with are really small). The issue with this is that you can't turn a sym array into a sparse matrix since matlab doesn't seem to allow that. does anyone have any suggestions?

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Stephen23 2016-8-1

编辑：Stephen23 2016-8-1

"some of the numbers I am dealing with are really small"

What order and how many digits ?

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

José-Luis 2016-8-1

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/297919-sparse-matrix-with-higher-than-double-precision#answer_230407

编辑：Thorsten 2016-8-1

The number of significant digits that you can store in a double is what limits things. As such, the smallness of the number doesn't really matter unless the values in your array span many orders of magnitudes and you need to keep more than (around) 15 digits.

If 15 significant digits is enough for you, you could always scale them, keep an index of the scaled numbers, and store as a double in a sparse array.

Please note that I don't have the symbolic math toolbox and I am taking you to your word that a sparse symbolic array cannot be generated.