# Felipe Yáñez

> Scientist and Principal Investigator at the Max Planck Institute for Neurobiology of Behavior. Uncovering the origins of neuronal diversity in cortex, and exploring human–AI teaming approaches for neuroscience.

## Identity

- **Name:** Felipe Yáñez
- **Site:** https://felipeyanez.github.io/
- **Email:** felipe.yanez@mpinb.mpg.de
- **Affiliation:** Max Planck Institute for Neurobiology of Behavior — https://mpinb.mpg.de/en/
- **GitHub:** https://github.com/felipeyanez
- **LinkedIn:** https://www.linkedin.com/in/felipe-yanez
- **Twitter:** https://twitter.com/yanezlang

## Background

Originally trained as an Electrical Engineer at Pontificia Universidad Católica de Chile (PUC), then specialized in Machine Learning at the Master MVA (Mathématiques, Vision, Apprentissage) in Paris. Worked in software engineering for two years before graduate school. Undergraduate internships at The ABB Group, MIT (RLE), and Inria Paris-Saclay (Parietal team). PhD from Eberhard Karls Universität Tübingen (2025).

## Research

Two ongoing inquiries:

### 1. Structural principles of the cortex

Cortical inhibitory neurons display enormous diversity in their cellular properties. Approach:
- **Dataset creation** — standardized and curated a large representative sample of cortical inhibitory neurons.
- **Robust classification** — assigned neurons into classes based on a highly accurate multi-modal assessment.
- **Flexible model development** — designed a self-supervised model to learn the statistics of a representative dataset across cortical layers.

Finding: depth-dependent gradients of cellular properties provide a link between them, suggesting an adjustment of cellular properties once integrated into local cortical circuits.

### 2. Human–AI collaboration for discovery neuroscience

As AI systems surpass human experts at predicting neuroscience experiment outcomes, what is the role for humans in the overall decision process? Effective team complementarity requires:
- **Calibrated confidence** — when an agent (human or AI) is more confident in their judgments, those judgments tend to be more accurate.
- **Classification diversity** — diversity holds when the errors in classification made by humans and AIs are not the same.

Where human judgment cannot be supplanted by AI, this opens the possibility for humans and AI systems to team together to make discoveries.

## PhD Thesis

- **On the classification of cortical inhibitory neurons** (2025, Eberhard Karls Universität Tübingen)
  Felipe Yáñez · https://dx.doi.org/10.15496/publikation-104703 · Public lecture: https://www.youtube.com/watch?v=yK4xWW_3z9E · Code: https://github.com/inhibicion

## Publications

- **Morphoelectric properties of inhibitory neurons shift gradually and regardless of cell type along the depth of the cerebral cortex** (2026, bioRxiv preprint)
  Felipe Yáñez†, Fernando Messore, Guanxiao Qi, Nima Dehghani, Hanno S. Meyer, Dirk Feldmeyer, Bert Sakmann, Marcel Oberlaender†
  https://doi.org/10.64898/2026.03.05.709819 · Code: https://github.com/inhibicion/decoupling

- **Confidence-weighted integration of human and machine judgments for superior decision-making** (2026, Patterns)
  Felipe Yáñez†, Xiaoliang Luo, Omar Valerio Minero, Bradley C. Love
  https://www.cell.com/patterns/fulltext/S2666-3899(25)00271-5 · Preprint: https://doi.org/10.48550/arXiv.2408.08083 · Code: https://github.com/braingpt-lovelab/haico

- **FOXP2-immunoreactive, corticothalamic pyramidal cells in neocortical layers 6a and 6b are tightly regulated by neuromodulatory systems** (2025, iScience)
  Guanxiao Qi, Danqing Yang, Fernando Messore, Arco Bast, Felipe Yáñez, Marcel Oberlaender, Dirk Feldmeyer
  https://doi.org/10.1016/j.isci.2024.111646

- **Large language models surpass human experts in predicting neuroscience results** (2024, Nature Human Behaviour)
  Xiaoliang Luo*, Akilles Rechardt*, Guangzhi Sun*, Kevin K. Nejad*, Felipe Yáñez*, Bati Yilmaz*, …, Bradley C. Love*
  https://doi.org/10.1038/s41562-024-02046-9 · Code: https://github.com/braingpt-lovelab/BrainBench · Site: https://braingpt.org/

- **Simulation-based inference for efficient identification of generative models in computational connectomics** (2023, PLOS Computational Biology)
  Jan Boelts, Philipp Harth, Richard Gao, Daniel Udvary, Felipe Yáñez, Daniel Baum, Hans-Christian Hege, Marcel Oberlaender, Jakob H. Macke
  https://doi.org/10.1371/journal.pcbi.1011406 · Code: https://github.com/mackelab/sbi-for-connectomics/

- **A hidden parameter network that explains topological properties observed in neocortex** (2021, ICLR Generalization Workshop)
  Felipe Yáñez, Daniel Udvary, Marcel Oberlaender

- **Primal-Dual Algorithms for Non-negative Matrix Factorization with the Kullback–Leibler Divergence** (2017, IEEE ICASSP, oral)
  Felipe Yáñez, Francis Bach
  https://doi.org/10.1109/ICASSP.2017.7952558 · Code: https://github.com/felipeyanez/nmf

- **Quantitative Susceptibility Map Reconstruction via a Total Generalized Variation regularization** (2013, PRNI)
  Felipe Yáñez, Audrey Fan, Berkin Bilgic, Carlos Milovic, Elfar Adalsteinsson, Pablo Irarrázaval
  https://doi.org/10.1109/PRNI.2013.59

†corresponding author · *major contribution

## Teaching & Outreach

- **Mentorship** — LatinX in AI (https://www.latinxinai.org/) and Deep Learning Indaba (https://deeplearningindaba.com/).
- **Fellowship** — Pillar VC (https://frequency.pillar.vc/) and Wilbe Science (https://app.wilbe.com/waitlist).
- **Teaching** — TA and lecturer at PUC Chile and Uni Frankfurt. Courses: Machine Learning, Pattern Recognition, Computational Neuroscience, Imaging, MRI Laboratory, Fluid Mechanics, Calculus, Differential Equations, Signals & Systems, Linear Algebra.
- **Undergrad project** — Volta UC: helped develop a 60 kW electric vehicle and produced the first technological documentary made in Chile. https://felipeyanez.github.io/undergrad/volta-uc.html

## Selected talks (recent)

- Cosyne 2025 (Montréal) — structural organization of inhibitory neurons across cortical areas and species
- PhD Defense 2025 (Tübingen) — depth-dependent variations reveal molecular identity of inhibitory neurons
- Congreso Futuro (Chile) — team collaboration with superintelligent AI systems
- SfN 2024 (Chicago) and invited talk at Northwestern Feinberg School of Medicine
- MIT McGovern (online) — molecular identity of cortical inhibitory neurons
- SfN 2023 (DC) and Johns Hopkins
- NWB Hackathons 2023 (Granada, Janelia)
- Mission KI · Deutsches Museum Bonn
- SfN 2022 (San Diego), FENS Forum 2022 (Paris), EPFL Neuro Symposium 2022
- ICLR 2021 Brain2AI Workshop
- SfN 2019 (Chicago) and Northwestern

## Topics

Machine Learning · Computational Neuroscience · Cortical Inhibitory Neurons · Human–AI Collaboration · AI for Science · Cell Type Classification · Connectomics · Simulation-Based Inference · Non-negative Matrix Factorization · MRI Reconstruction