As the author noted after you posted this, it’s not a given that the ML algorithm is O(n). It may be constant time (by looking only at column headers and a samp

As the author noted after you posted this, it’s not a given that the ML algorithm is O(n). It may be constant time (by looking only at column headers and a sample of data, say). That said, I was really more interested in practical runtimes. Like, in practice the ML may have a high startup cost (e.g., due to cost of loading a model), whereas for most sized datasets linear complexity may be fine...