[conference] Presented at NeurIPS 2025 Data on the Brain & Mind Workshop
I presented our work on fast reconstruction of degenerate neuron populations at the NeurIPS 2025 workshop in San Diego, CA!
I’m excited to share that I presented a poster at the Data on the Brain & Mind Workshop at NeurIPS 2025 in San Diego, CA!
The work
The poster presented our method for reconstructing degenerate populations of conductance-based neuron models directly from spike times using deep learning - the same work from my first preprint that I shared earlier this year.
Key idea: We combine deep learning with Dynamic Input Conductances (DICs) to solve a fundamental inverse problem in neuroscience. Starting from just spike times (the most widely available experimental data), our pipeline:
- Maps spike times → DICs using a lightweight neural network
- Generates degenerate populations via an iterative compensation algorithm
- Produces diverse yet functionally equivalent models within milliseconds
The method handles realistic variability (including Poisson spike trains), works across different neuron types, and comes with open-source software including a GUI for experimentalists.
Presenting at the Data on the Brain & Mind workshop at NeurIPS 2025
Why it matters
Spike times tell us what neurons do, but not which combinations of ion channels produce that activity. Our pipeline bridges this gap by:
- Enabling inference from the most accessible experimental data
- Capturing neuronal degeneracy (many solutions → same behavior)
- Providing interpretable intermediates (DICs) that explain excitability
- Running fast enough for real-time applications
Resources
📄 Full paper: Available on ORBi
📝 Blog post: Read about my first preprint on this work
💻 Code & Software: GitHub repository
🎓 Workshop: Data on the Brain & Mind at NeurIPS 2025
The workshop brought together researchers working at the intersection of neuroscience, machine learning, and data analysis. It was fantastic to get feedback and discuss how this approach could be applied to experimental data and extended to network-level inference.
Thanks to everyone who stopped by the poster and to my co-authors Damien Ernst, Guillaume Drion, and Arthur Fyon for their collaboration on this work!
Contact: Questions about the method or the software? Reach out at jbrandoit@uliege.be