Neuromatch Academy: a 3-week, online summer school in computational neuroscience

Bernard Marius ’t Hart, Titipat Achakulvisut, Athena Akrami, Bradly Alicea, Ulrik Beierholm, Gunnar Blohm, Kathryn Bonnen, John S. Butler, Brandon Caie, You Cheng, Hiu Mei Chow, Isaac David, Eric DeWitt, Jan Drugowitsch, Kshitij Dwivedi, Pierre-Étienne Fiquet, Jérémy Forest, Byron Galbraith, Qinglong Gu, Pankaj Gupta, Alexandre Hyafil, Konrad P. Körding, Arvind Kumar, Patrick Mineault, John D. Murray, Megan A. K. Peters, Paul Schrater, Carsen Stringer, Pascal Wallisch, Brad Wyble2021 · DOI

Abstract

Neuromatch Academy (https://neuromatch.io/academy) was designed as an online summer school to cover the basics of computational neuroscience in three weeks. The materials cover dominant and emerging computational neuroscience tools, how they complement one another, and specifically focus on how they can help us to better understand how the brain functions. An original component of the materials is its focus on modeling choices, i.e. how do we choose the right approach, how do we build models, and how can we evaluate models to determine if they provide real (meaningful) insight. This meta-modeling component of the instructional materials asks what questions can be answered by different techniques, and how to apply them meaningfully to get insight about brain function.

Introduction

Computational neuroscience is a highly interdisciplinary field, sitting at the intersection of biology, physics, mathematics, and computer science. Despite its importance, high-level training in this field is often concentrated in elite, well-funded institutions, creating a "geographical lottery" for aspiring researchers. Furthermore, traditional education often emphasizes the mechanics of mathematical modeling without addressing the underlying philosophy: how to choose the right model for a biological question. Neuromatch Academy (NMA) was established to address these gaps by providing a globally accessible, high-quality curriculum focused on "meta-modeling."

Curriculum and Pedagogy

The NMA curriculum spans 15 days of intensive instruction, preceded by optional Python preparation. The program utilizes a "flipped classroom" model where students engage with pre-recorded lectures and hands-on Python tutorials in Jupyter notebooks. A critical innovation is the small-group learning environment, where students are organized into pods of approximately ten, each led by a dedicated Teaching Assistant (TA). This structure ensures personalized support and social accountability, which are often missing in traditional Massive Open Online Courses (MOOCs).

The 3-week curriculum structure of Neuromatch Academy, covering model types, machine learning, dynamical systems, and deep learning.

The "meta-modeling" framework is central to the pedagogy. Rather than treating models as "black boxes," the course explicitly teaches students how to evaluate whether a model provides meaningful insight into brain function. This involves a rigorous quality control process for all materials, ensuring pedagogical consistency across the diverse topics covered.

Real-World Data and Group Projects

To bridge the gap between theory and practice, NMA incorporates five curated, high-quality datasets for student group projects. These datasets represent a range of modalities, from calcium imaging and Neuropixels recordings in mice to human fMRI and behavioral data.

Summary of curated datasets and potential project applications provided to students for group projects.

By working with "messy" real-world data, students learn to navigate the complexities of data analysis and model fitting in a research context. This hands-on experience is vital for developing the intuition necessary to apply computational tools to biological questions.

Global Impact and Results

In its inaugural 2020 session, NMA achieved unprecedented scale for a specialized scientific course. The program reached nearly 8,000 learners across 64 countries, including 1,757 interactive students and approximately 6,000 observers. Unlike typical MOOCs, which often see single-digit completion rates, 86% of NMA’s interactive students completed at least half of the curriculum. Student satisfaction was remarkably high, with 94% of respondents stating they would recommend the program to others.

Challenges and Future Directions

While successful, NMA faces ongoing challenges. The three-week format is extremely demanding, requiring significant prerequisites in mathematics and neuroscience. Additionally, while the program improved global access, achieving full inclusivity across gender and race remains a priority. The sustainability of such a massive volunteer-led effort is also a key consideration for future iterations.

Conclusion

Neuromatch Academy demonstrates that high-level scientific education can be scaled globally through decentralized mentorship and open-source materials. By focusing on the logic of modeling and providing access to real-world datasets, NMA empowers a new generation of computational neuroscientists regardless of their geographic location.

Summary updated

Abstract

Neuromatch Academy (https://neuromatch.io/academy) was designed as an online summer school to cover the basics of computational neuroscience in three weeks. The materials cover dominant and emerging computational neuroscience tools, how they complement one another, and specifically focus on how they can help us to better understand how the brain functions. An original component of the materials is its focus on modeling choices, i.e. how do we choose the right approach, how do we build models, and how can we evaluate models to determine if they provide real (meaningful) insight. This meta-modeling component of the instructional materials asks what questions can be answered by different techniques, and how to apply them meaningfully to get insight about brain function.
BibTeX
@misc{hart2021neuromatch,
  author = {Bernard Marius ’t Hart and Titipat Achakulvisut and Athena Akrami and Bradly Alicea and Ulrik Beierholm and Gunnar Blohm and Kathryn Bonnen and John S. Butler and Brandon Caie and You Cheng and Hiu Mei Chow and Isaac David and Eric DeWitt and Jan Drugowitsch and Kshitij Dwivedi and Pierre-Étienne Fiquet and Jérémy Forest and Byron Galbraith and Qinglong Gu and Pankaj Gupta and Alexandre Hyafil and Konrad P. Körding and Arvind Kumar and Patrick Mineault and John D. Murray and Megan A. K. Peters and Paul Schrater and Carsen Stringer and Pascal Wallisch and Brad Wyble},
  title = {{Neuromatch Academy: a 3-week, online summer school in computational neuroscience}},
  year = {2021},
  doi = {10.31219/osf.io/9fp4v},
}
Figure