Baseline meaning in machine learning competitions In some shared tasks "competitions" that are specialized in machine learning .. when they reveal the results of the tasks, they added a row in the results called "Baseline - as in the attached image" .. what is the meaning of it ?!

 A: I've worked with Kaggle before to setup ML competitions at a company I used to work for.
In our case, the baseline was the score from a hastily thrown together minimal viable model for the task in the competition.  In other cases, where there was possibly considerable prior art, it may be the score of the company's current production solution.  In either case, it is a score that the company feels should be beat by any model worth considering.
The instructions to the competition should say something about the baseline model.
A: Usually in these types of competitions, there have already been attempts to solve the problem before it's crowdsourced. The "baseline" metric will represent the best solution found at the time the competition is opened. In the Netflix Prize for example, the goal was to improve over Netflix's own Cinematch algorithm, which would be listed as the baseline here. Anything that falls below the baseline in terms of performance is not an improvement over pre-existing methods. Exactly what that baseline method is will of course vary between problems.
A: It is the score achieved by the organizers of the competition using some approach of their choice, that serves as some kind of benchmark (usually suggests that your solution is poor if it is below it). It can be something that was achieved using very simple methods, some kind of “standard” approach, using the method that is used by organizers on daily basis with this data, current state of art approach, or basically anything.
