Greetings from Taiwan :) From the challenge description, we will be submitting trainable models that are run on servers provided by the competition. I was wondering about the details of the computational resources provided since this is likely a key factor, specifically: (1) Is there a limit to size of model? (2) The computing power of computing instances? Surely GPU is provided? (3) How many models can a person submit in parrallel? How to prevent people from submitting through several accounts to gain advantage? The training set is huge, so I'm very thankful for this contest.Finally such challenges are not only imitted to the privileged few!! Cheers, Sanders

Created by sanders lin santo0520
Hi bnorth, We are still discussing the policy on how we will serve the data, however for now you can expect to be able to uniquely identify an exam (and its data) across multiple runs.
As a follow-on: Will the training images always be served in the same order? For example, if I wanted to only train on the first 50k images, to decrease training time, will I get the exact same images each time I request the first 50k? Thanks!
Hi, I'll be a little short on specifics, but will try to answer anyway. First, both IBM and Amazon have generously agreed to provide some very high powered GPU compute nodes for the challenge. The training time and capacity must be kept within limits, but our goal is to provide sufficient resources to let the better methods shine. We're currently testing dry-run models to determine what those limits should be. It's a tricky balance to stick within our allotments, yet enable widest participation. We hope for a cooperative spirit from the participants in that respect. Running submissions in parallel will likely be discouraged and will, in any case, deplete your quotas quickly and without benefit of feedback from previous runs. We have measures in place to detect and discourage unfair behavior. Also, we encourage everyone involved to compete in the challenge with the goal of helping cancer patients in mind.
Dear Sanders, I apologize for the delay in response. Unfortunately, I do not know the answers to your questions, but I will let the other challenge organizers know so that they can answer your questions. Thank you for your patience. Best, Thomas

Regarding computational resources page is loading…