Matrix Row / column wise min/max

Question

Matrix Row / column wise min/max

sinhrks opened this issue 8 years ago · comments

It is convenient if we have row / column-wise .min and .max of Matrix.

accept Axes to specify axis like other aggregation funcs.
min / max logic should be the same as argmin / argmax. (thus requires PartialEq)

James Lucas · Answer 1 · Thu Dec 22 2016 22:08:06 GMT+0800 (China Standard Time)

I agree that this would be a useful addition. The only issue is that we would require the PartialEq trait and so we cannot add this to the BaseMatrix trait.

In order to implement this properly we would need column iterators ( #3 ) which are on the way with #84 .

James Lucas · Answer 2 · Fri Dec 23 2016 02:14:14 GMT+0800 (China Standard Time)

Excuse part of my above comment - we can add this to the BaseMatrix trait by adding a where clause to the function which requires T: PartialEq.

In which case I think this is a solid addition and I would welcome a PR. If you want to work on this feel free to implement the function without column iterators for now. We should use the Axis enum to keep things tidy.

Andreas Borgen Longva · Answer 3 · Fri Dec 23 2016 04:07:02 GMT+0800 (China Standard Time)

Also note that - for the sake of performance - the column-wise min/max is probably better implemented by using the row iterator and folding the rows by accumulating min/max values for each column entry in the row into a vector, rather than using column iterators. This because it's likely far more cache-friendly.

That said, measure first and all that.