Portrait format. Max size: 20 inches wide and 30 inches tall.
We have about 50 posters at each session, so PLEASE keep your posters within this size so everyone fits into our assigned room.
Note that this is different than the suggested poster size for the main NIPS conference.
Poster Session Assignments
Please check our accepted posters page for the assigned time of your poster presentation at the conference, as well as instructions about when to setup, present, and take down your posters.
Suggested Tips for an effective poster at NIPS ML4H
- Succinctly state the machine learning / health problem and its significance. Why is this problem a high value application of machine learning? What is the problem’s clinical or scientific relevance?
- Show important equations and mathematics surrounding your work. Include the key mathematical ideas around your machine learning representation. What is novel about your representation?
- Include a description of the data set you are working with and include exact sample sizes for data splits (e.g. train/test) when possible as well as relevant data collection details.
- Use system diagrams and abstraction concepts to visually show your representation where applicable. Use appropriate software engineering practices in designing your system diagrams, consulting your favorite software engineering / design patterns textbook.
- Please include implementation details of your setup including what hardware or high performance computing cluster service was used to train / test your machine learning model (if applicable). Please include discussion of the relevant software platforms or Application Programming Interfaces (APIs) that were used to build your model.
- At NIPS ML4H, we also celebrate novel hardware or software that advances deep learning / machine learning for health applications. Please discuss any custom implementations that you did in pursuit of your research (if applicable).
- Include discussion of all statistical modeling methodology used on the data set as well as exact definitions of evaluation metrics.
- Provide detailed explanation of your results and how you chose to evaluate your models. It is helpful to explain why you chose particular architectures, features, or models over alternative designs, what criteria were used, and why these are good criteria for the problem you studied.
- Include conclusions drawn from the scientific study and suggestions for future work.
Tips for Figures / Illustrations
- Figures should be readable from 3' (90cm) away.
- Use legible and visible fonts for plot figures and legends (generally 1" or 2.54 cm high).
- It is helpful to print a proof before printing the actual poster to assess relative sizing of titles / text.
- Refrain from adding too much clutter to figures. Ensure that figures are readable.
- Simple, interpretable figures (even if your model is complex) along with well-designed metrics, facilitates the most concrete discussion.
Tips for System Diagrams
- Show the full flow of data inputs and outputs in the system.
- It is helpful to show modular figures when applicable..
Tips for Result Tables
- Clear headings and numbering of tables can help discussion of results.
- Adding colors to tables is useful highlight particular results of interest.
- When working with human subjects, please include acknowledgements, IRBs, any required consent statements, disclosures, and disclose all relevant data collaborations.