Currently np.argmax, np.argmin, np.max and np.min return the maximum and minimum values respectively.
numpy.argmax(a, axis=None, out=None)
Often times we need to get the topKmax or topKmin values, mostly in the case of machine learning workloads where you need top K classes for a given input.
Requesting support for these scenarios in numpy. This scenario could be supported by np.max (and other functions), or there could be a separate family of methods np.maxK, minK etc?