Merge pull request #52 from esl-epfl/pr-workshop

Reproducibility workshop

website/content/workshop/index.md

@@ -0,0 +1,67 @@
+---
+title: "Reproducibility Workshop"
+description: "A workshop aimed at investigating the reproducibility of seizure detection models."
+---
+
+We have all faced reproducibility challenges in machine learning, particularly when applying complex models like those used for epileptic seizure detection. You read an intriguing paper or model description, eagerly anticipating the release of source code and datasets, but then, either the resources are incomplete or outdated or, worse, they simply don't work. This is not only frustrating but also limits progress in the field, particularly when it comes to applying machine learning algorithms in real-world clinical settings, like epilepsy monitoring units.
+
+In this workshop, we want to explore this issue by investigating the reproducibility of algorithms used in epileptic seizure detection. The results will help assess the current state of reproducibility within this area of research and uncover common pitfalls faced when attempting to replicate results.
+
+While reproducibility in research can be ensured without code if detailed methodology is provided, the availability of well-documented code plays a critical role in ensuring others can reproduce the results efficiently. We'll explore how to assess reproducibility better, improve accountability, and foster best practices in sharing code and data.
+
+## Reproducibility workshop
+
+The workshop will focus on algorithms for automated patient-independent seizure detection algorithms that have been published in the last five years (>2020) and that have source code accompanying their publication.
+
+We aim to investigate the following:
+
+1. **Stage 0 (Data Extraction):** Review all patient-independent seizure detection algorithms published in the last five years, focusing on papers that make models or source code publicly available. We will exclude papers that train models on private data that prohibit reproduction. We will investigate whether these models or codes are easy to run or reproduce.
+2. **Stage 1 (Reproducibility Hackathon):** This hands-on hybrid hackathon will bring together researchers, students, and industry professionals to reproduce the results of a few selected seizure detection models. Participants will work on reproducing a subset of models, evaluating how easy it is to implement and whether any obstacles arise.
+3. **Stage 2 (Scalable Reproducibility with SzCORE):** This stage will follow up on the workshop by evaluating models on independent datasets using the [SzCORE methodology]({{< ref "/framework" >}}). This will ensure results are comparable between publication that could be reproduced.
+
+## Why
+
+The impact of this project is far-reaching:
+
+- **Highlight strengths and weaknesses** of seizure detection models regarding reproducibility.
+- **Identify best practices** for code and data sharing to improve reproducibility and foster accountability.
+- **Encourage transparency** in machine learning research, enabling clinical adoption of seizure detection algorithms.
+- **Allow comparisons** in seizure detection algorithms by comparing them on common datasets with unified performance metrics.
+- **Offer students and early career researchers** a hands-on opportunity to engage in cutting-edge research and contribute to the community’s understanding of reproducibility in machine learning models.
+- And most importantly: **It will be fun and educational**!
+
+## Who Can Participate
+
+This workshop is open to master students, PhD students, postdocs, and researchers in machine learning, neuroscience, or clinical research who are interested in exploring reproducibility issues in epileptic seizure detection models. Participants should be familiar with machine learning methods and EEG data analysis.
+
+## Where
+
+The hackathon will take place in a hybrid format:
+
+- **In-person** TODO set venue.
+- **Virtually** for those who are unable to travel or prefer to participate remotely.
+
+## When
+
+- **Registration Deadline:** Q1, 2025
+- **Decision on Participation:** XXX, 2025
+- **Workshop Dates:** Q1-Q2, 2025
+
+## How to Apply
+
+Please fill out the registration form [here](#). Participants will be selected based on their background and motivation, and we will notify successful applicants by XX, 2025.
+
+---
+
+## Benefits of Participation
+
+Participants in this workshop will gain:
+
+- **Insights into reproducibility**: Understand how reproducibility issues affect machine learning models for clinical applications.
+- **Hands-on experience**: Work directly with seizure detection models and datasets, solving real-world challenges in reproducibility.
+- **Collaboration**: Network and collaborate with like-minded researchers, clinicians, and engineers in the epilepsy and machine learning fields.
+- **Educational opportunities**: Learn about best practices in reproducible research, code sharing, and dataset standardization.
+
+---
+
+We look forward to collaborating with you to improve reproducibility in epileptic seizure detection!