Both Proceedings papers and Findings papers should be anonymous at the time of submission (no identifying information present in the submitted manuscript).
Proceedings Paper
Excellent ML4H proceedings papers should be compelling, cohesive works with a high degree of technical sophistication as well as clear and high-impact relevance to health. Demonstrating each of these qualities in your work will be essential to acceptance. Submissions to the proceedings track cannot already be published or under review in any other archival venue. Papers accepted to the proceedings track are archival and will appear in the ML4H proceedings. Reviewers will be asked to answer specific questions regarding technical sophistication and relevance to health in the review form, and authors may benefit from explicitly discussing these points in their papers.
Technical Sophistication
Technical sophistication is required for acceptance for an ML4H proceedings track paper. Merely applying well-established state-of-the-art techniques to a health dataset and demonstrating good performance is insufficient. Although “technical sophistication” is a broad term, we highlight two critical aspects here: novelty and rigor.
An excellent ML4H proceedings paper will demonstrate novelty. Technical novelty comes in many forms, but some type of novelty or innovation in method design, construction, evaluation, or use is required.
Rigor is also critical for a high-quality ML4H paper. Rigor will mean different things for different works, but, largely, rigor implies that the conclusions presented are well supported by empirical evidence. Usually, this involves appropriate statistical techniques for model design and evaluation, including creating separate train, validation and completely held-out test sets. The paper should also investigate different configurations of the proposed system.
Relevance to Healthcare
Health data and systems pose many unique challenges that often warrant novel techniques from machine learning or data science to address. An appreciation of these challenges and a focus on health applications, even if actual deployment is not yet feasible, should be demonstrated in your paper. While we especially encourage submissions relevant to this year’s themes, any submission that falls under the purview of machine learning for health will be considered.
Exemplar papers
These proceedings papers from ML4H held in prior years demonstrate both technical sophistication and relevance to healthcare and present their ideas in a clear and succinct manner.
- Counterfactual and Factual Reasoning over Hypergraphs for Interpretable Clinical Predictions on EHR
- Predicting Treatment Adherence of Tuberculosis Patients at Scale
- End-to-End Sequential Sampling and Reconstruction for MRI
- Early Exit Ensembles for Uncertainty Quantification
For additional tips on how to write an excellent technical paper, check out Zach Lipton's blog on writing good ML papers and these guidelines on how to write a good NeurIPS paper.
Findings Paper
An excellent Findings paper is one that leads to insight at the symposium through interaction with other attendees. This can be through presenting new ideas/ways of thinking, leading to insightful discussion and feedback, dissemination of new valuable resources, or enabling new opportunities for collaborations. We allow submissions to the Findings track that are under review at another venue (e.g., conferences, workshops, journals) as long as you adhere to the other venue’s submission policies. Upon acceptance, Findings papers are non-archival and will not appear in the ML4H proceedings, but we invite (but do not require) authors to submit their findings to the ML4H arxiv.org index.
Submissions to the Findings track (previously known as the Extended Abstracts track) should demonstrate that the work will produce fruitful discussion when presented at the symposium. Highlight opportunities for insightful discussion and demonstrate that your work will contribute to a creative, engaging, and constructive poster session. Reviewers will be explicitly asked to gauge how valuable they feel this work could be to other attendees of the event, as well as how valuable attending the event could be to the author of the abstract. This is not a license to submit low quality or barely begun work -- while these submissions may garner constructive comments during the review process, they will not likely generate useful discussions or insightful feedback during the event.
Reviewers will be asked to answer specific questions regarding the relevance to health in the review form. It may be beneficial for authors to explicitly discuss these points in their submissions.
Exemplar Findings
These abstracts from a prior year’s event presented preliminary, but promising ideas that served as good discussion points between the authors and other attendees.
- Extending the Neural Additive Model for Survival Analysis with EHR Data
- Language Model Classifier Aligns Better with Physician Word Sensitivity than XGBoost on Readmission Prediction
- Transfusion: Understanding Transfer Learning for Medical Imaging (This is the submitted, extended abstract version -- the work has also been published separately as a full paper, and interested readers should see that here)