Algorithms


An algorithm on grand-challenge.org is a machine learning solution to a medical problem. On grand-challenge.org, algorithms are hosted as container images that are executed on a set of defined inputs and that produce a set of outputs. The inputs and outputs can be text, numbers, booleans, medical images, or annotations. You can also have multiple inputs and outputs for an algorithm.

Hosting an algorithm on grand-challenge.org is an easy, secure, and convenient way to publically advertise your machine learning solution for a certain medical problem. People can easily try your algorithm out by uploading an image and inspecting the result in one of our integrated viewers. You can also link your algorithm to an archive on Grand Challenge and have the algorithm run on all images in the archive asynchronously. Finally, algorithms on grand-challenge.org can also be used as submissions for challenges, for more on that see here.

There are two types of users for algorithms:

  • Editor: An editor is someone who can edit the algorithm settings, add other editors, add and remove users, and update the algorithm's container image. An algorithm can have as many editors as you want.
  • User: A user is someone who can try out the algorithm by uploading an image to the algorithm and inspecting its output.

Public vs. private algorithms

An algorithm can be public or private. If an algorithm is private, it will not appear in the list of algorithms. Users can then only request access and use the algorithm, if given the url. A public algorithm in turn is visible on the algorithm overview page. Even if an algorithm is public, users first need to request access before being able to try the algorithm out.

Credit system

An algorithm editor can choose to ask for credits from users who want to try out the algorithm. These can be set by the editor in the algorithm settings. A credit of 0 means users can run the algorithm an unlimited number of times. If a credit value above 0 is set, these will be deducted from the users' credits whenever they submit a file to run the algorithm on. A user has 1000 credits per month at their disposal.