Skip to content
This repository has been archived by the owner on Jun 22, 2022. It is now read-only.

User Guide

kamil-kaczmarek edited this page Dec 15, 2017 · 25 revisions

How to work with Minerva?

Problem

As a User, you start working on a problem, by opening the welcome notebook, for example, Right Whales Recognition or Fashion MNIST.

Each problem is a complex, end-to-end pipeline that leads from raw data to the final score. Minerva developers have implemented entire pipeline and divided it into logical pieces. Such pieces are called tasks - User's main point of interest.

Task

Task is designed in such a way, that it contains single processing step, such as ConvNet implementation or post-processing function. It is presented in its own notebook, where User can find the description of the task, learning objectives, what is expected as a solution and some auxiliary information that should help in the implementation process. Note, that User may work on tasks in any order.

Technically, User works on task by implementing the solution function, which stub is already in the notebook. When the solution is ready, User submits the task, and the corresponding part of the pipeline is replaced by the User's implementation. Recall, that entire pipeline is pre-implemented by the Minerva team.

User submits the task by running a single command. The structure of this command is explained in detail in the Welcome notebook (again Right Whales Recognition or Fashion MNIST).

Task score

Currently, feedback is provided in a binary way: if the final score is very close to the default pipeline implementation, User receives appropriate praise, if not, User is asked to try again.

Towards entire problem re-implementation

Problem consist of many tasks and User is encouraged to solve all of them. User can work on Tasks in any order and eventually re-implement complex data scientific pipeline.

Clone this wiki locally