| |
|
| |
| DETAILS |
|
Sorting might sound like an undergraduate topic, but it turns out you can literally do Machine Learning by sorting data. This is the result of a theorem I proved in my paper, "Sorting, Information, & Recursion", in which I prove that a list is sorted if & only if every entry is adjacent to its Nearest Neighbor.
We'll walk through the paper as a whole, which also presents what I believe to be the fastest known sorting algorithm, with a runtime of O(log(N)).
This means that you can do Machine Learning with an O(log(N)) algorithm, which is pretty impressive.
These techniques were implemented in the "Massive" version of my AutoML software BlackTree, which can train & test 500,000 rows in about 10 minutes, running on an ordinary PC.
As you can tell, this is more than an academic result, it produced extremely powerful commercial software, which we will discuss as well.
See you all then!
Charles
|
|
|
|
|
|
|
|
