Harvard Catalyst is exploring the value of crowdsourcing approaches to solving algorithmic challenges with health-related applications. This approach provides the means to leverage very highly-skilled algorithm and software developers from around the world, engaging them in a competition aimed at solving a submitted problem for an expressed award; e.g. prize money.
Winning algorithms tend to outperform solutions derived through conventional means, sometimes by orders of magnitude. In a computational immunogenomics challenge sponsored by Harvard Medical School, the solutions derived from crowdsourcing outperformed the best academic effort by a factor of 100. Moreover, these solutions were derived within two weeks, and several new algorithmic approaches to solving the problem were also discovered.
Harvard Catalyst would like to invite all members of the Harvard community to consider submitting a health-related (basic science with health-related application/preclinical/clinical) computational or algorithmic problem to be solved via crowdsourcing. This approach leverages very high-skilled algorithm and software developers from around the world to compete to solve your problem for prize money. Winning algorithms tend to outperform solutions derived through conventional means, sometimes by orders of magnitude.
Recently, a test challenge in the computational immunogenomics area, sponsored by Harvard Medical School, obtained more than 600 software submissions from over 125 individuals. The best solutions outperformed the best academic effort by a factor 100. Moreover, these solutions were derived within two weeks, and several new algorithmic approaches to solving the problem were also discovered.
Harvard Catalyst is interested in further exploring the value of contest-based crowdsourcing approaches to solving computational algorithmic challenges and has funds available to provide prize money for a number of challenges to be run on the TopCoder crowdsourcing platform. Harvard Catalyst and TopCoder will provide the support and infrastructure to help develop the problem statement so that it can be ready for competition. Note that the submission itself is very simple and undetailed.
Harvard Catalyst invites faculty, associates, students, and other members of Harvard University and its affiliated institutions to consider submitting a computational or algorithmic problem with health-related applications to be solved through crowdsourcing.
Funds available to support crowdsourcing contests are limited and therefore not all submissions will be able to be pursued. Factors that will be considered in selecting challenges include: 1) suitability of the question, 2) availability of sufficient data or information to support the efforts of the solvers, 3) availability of a validation dataset, 4) evidence that the algorithmic problem is important to the success of the submitter's research or progress in a particular field, 5) the relationship of the problem to important classes of algorithmic problems, and 6) congruence with the overarching mission of Harvard Catalyst.
Figure 1: Overview
A problem statement, test data, and a scoring algorithm are used to issue a challenge on the crowdsourcing platform (TopCoder) with prize money funded by Harvard Catalyst. Harvard Catalyst will work with the researcher to effectively package the submission (i.e., problem statement, test data, and scoring algorithm) for delivery to TopCoder. At the end of the challenge, the solution algorithm is transferred to the originating researcher. (See a sample problem statement used in the previous HMS crowdsourcing challenge.)
About This Research Program
This program is funded by a grant awarded by NIH/NCRR that supports Harvard Catalyst's ongoing investigation of the applicability and results of open innovation methodologies in academic biomedical research.
Eva Guinan, MD, HMS & DFCI, Harvard Catalyst
Karim R. Lakhani, PhD, Harvard Business School
Patrck Gaule, PhD, HMS & Harvard Catalyst
Please E-mail Patrick Gaule if you have any further questions.