Introduction

This repository contains the solution to same-named Kaggle competition.

Key to solution

Basically, the concept is in to train not only one model at time, but two: first will be training to precise bot and second - human.

It can be strange with the first look: binary classification asks for one model which can divide dataset on two parts, but our taks is not truly about that: the task says to calculate probabilities, when it becomes clear, current approach cease to seem such strange.