Mindsync platform Competitions

Mindsync
mindsync.ai
Published in
2 min readApr 17, 2019

Competitions are a way to get a strong solution to a problem in a relatively short time. The advantage of the competition is that a large number of competent people work on the task, who manage to test a large number of hypotheses. As a result, the customer receives a working algorithm and, what is often more important, the knowledge that the result obtained cannot be improved at low cost.

In other words, the decisions obtained as a result of contests are an approximation to the upper possible limit of the quality of the ML model.

This circumstance makes competitions extremely effective in applications where even a relatively small improvement in the quality of the forecast is in demand by the business.

As example, to get a result similar to the competition, and to check a comparable number of hypotheses, you need several (tens) personyears of work.

Task requirements come from the customer, but at this stage it’s possible to involve community experts for consultation and assistance.

The competition mechanics are tailored to a specific task, but the general mechanism is as follows:

1. The terms and metric are determined and the data for training is given to the participants.

2. Participants optimize the metric in the framework of the contest rules.

3. The final assessment of the ML-models takes place on the test dataset.

4. Winners and top participants are awarded in proportion to their contribution.

5. Solutions, approaches and experiences are documented and become knowledge of the community.

6. All participants receive ratings.

7. The ML-model is published on the platform market.

The mechanics of the competition, additional restrictions, the method of choosing the winner and other nuances may, if necessary, be redefined by the Customer to match the original task.

The benefits of using the MindSync platform:

• For participants — experience, knowledge, contribution to the community, reward, rating.

• For customers — the solution of their problem, insights about their specific data.

• For the platform — community development, increasing engagement, developing market models.

--

--