How to change Matlab default data precision other than 'vpa'?

I would like to perform a matrix division on two large matrices, A \ b, (e.g., 1000-by-1000 and 1000-by-1). I always get a warning from this operation:Warning: Matrix is close to singular or badly scaled. Results may be inaccurate. RCOND = XXX

I think the problem is mainly because some elements in A and b have more than 16 digits.

I tried using 'vpa' to increase the precision, but it is super slow for converting a large matrix and the data type is no longer 'double' which would cause some problems later in my algorithm.

Could you please tell me if there is any way to change Matlab default data precision other than using vpa?

(P.S. you might suggest me using Levenberg–Marquardt algorithm to tackle the problem, but I found only a very large damping parameter can remove this warning, which is not very ideal in my case)

1

Best Answer

There is no such thing as 'too many digits' in Matlab matrices. They are stored as 8 byte doubles. Your assumption about the problem is wrong.

Random Posts

How to find out if a directory is a valid PostgreSQL clusterHow do I use Get-MPComputerStatus from Remote PC?Polymorphism in GraphQL with a one-to-many relationshipsqlalchemy simple example of `sum`, `average`, `min`, `max`How to rewrite shared Git history already on Github?Object type in mongooseRemove last item from arrayPython empty constructorOpen Group Policy Management Console (gpmc) for Another DomainVIM Select Entire LineHow to import an fbx in Unity correctly?Azure Devops 203 Non-Authoritative Information with REST APIwhat do the "-o" "-f" "-c" and "-g" etc mean in make?HttpURLConnection timeout settingshow to add Javascript code in specific page in SquareSpace websiteconsole.log timestamps in ChromeLogical XOR operator in C++?How to drop a table if it exists?python recursive pascal triangleHow to build a simple RSS reader in Python 3.7?