High Dimensional Matrix

KNIME Analytics Platform
1

In our applications, we need to handle a lot of 3D or 4D matrices. I am new to KNIME, and may I ask if there is any node can handle this kind of high dimensional matrices? If I created a R snippet node to load 3D matrix, how do I pass it to the next node (such as an R, or Python) to process?

Any advice would be appreciated.

 

 

2

Maybe pivoting can help? Although it can loose some type information.

I am not sure how would you like to process the information further though.

I mean I think you can generate a table like this:

Column (optional) Dimension1 Dimension2 Dimension3 Value
matrix1 0 0 0 3.4
matrix1 1 0 0 3.25
...        
matrixx n m o 1.442
3

Thanks for your reply aborg! The problem is that my matrices are potentially very large.

For example, a typical one can be 1000x1000x2000 in size, and each element is, a float or double. Not only I want to save and transfer it along the workflow, but also I would like to process it. For example, a 3D Gassian filer to convolve the data.

I noticed that in the recently released KNNIME Image Processing package, it possible be able to load a 3D image? Is there a way that I can using R/Python to load my 3D matrix, and convert it to a 3D image that can be processed by the Image Processing nodes?

 

4

I think the BioFormats library is able to handle multiple dimensional images. As I know, the KNIP extension uses that, so you might be lucky if it fits your use case. I would give it a try.

5

Thanks for your info! I will take a look.

6