Question

I’m afraid that I can’t describe the problem so I draw a sketch of it.Anyway,what I need is to find the max values along the 0th axis in a numpy ndarray,i.e.array.shape(5,5,3), and their corresponding "layer numbers", and use the "layer numbers" to create a new 2d array with shape of (1,5,3).Hope I’m giving a clear description here..thanks a lot.

Answer 1

If you check the documentation of np.max, you’ll see it takes an axis argument:

a.max(axis=0)

But that won’t help you yet. However, there’s a function argmax that gives you the indices of the maxima along a given axis:

a.argmax(axis=...)

So, let’s find your first (5,5) array: it’s a[...,0]. You can find the position of the maxima per rows (or columns) with a[...,0].max(axis=1) (or 0), and use that to find the values on the other sides.