Process to removes a column from INDArray

ND4J
1

Hi all,
I am looking for a performant way to remove a single column from an INDArray, which I can use in an iterative fashion.

To provide more color, I have a Java process which operates on an INDArray of several hundred thousand columns and several thousand rows. My analysis requires me to remove a single column and operate on the remaining matrix.

For example, if I have a 4 row x 3 column matrix,

  • remove column 0 only - recalc,
  • then remove column 1 only - recalc,
  • finally remove column 2 - recalc.
    Noting that each recalc operation above was on a 4 row x 2 column matrix.

I’m currently using originalArray.get (NDArrayIndex.interval…) to determine the left and right, and hstack-ing them, but this is quite slow as the INDArray size increases.

Thank you!