Martin Schrimpf
+41 21 693 12 53
Office:
AI 2241
EPFL
>
SV
>
INX-SV
>
UPSCHRIMPF1
+41 21 693 12 53
Office:
AI 2241
EPFL
>
IC
>
IINFCOM
>
UPSCHRIMPF2
Web site: Web site: https://sin.epfl.ch
Web site: Web site: https://ssc.epfl.ch
Biography
Martin's research focuses on a computational understanding of the neural mechanisms underlying natural intelligence in vision and language. To achieve this goal, he bridges Deep Learning, Neuroscience, and Cognitive Science, building artificial neural network models that match the brain’s neural representations in their internal processing and are aligned to human behavior in their outputs.He completed his PhD at the MIT Brain and Cognitive Sciences department advised by Jim DiCarlo with collaborations with Ev Fedorenko and Josh Tenenbaum, following Bachelor’s and Master’s degrees in computer science at TUM, LMU, and UNA. His previous work includes research in human-like vision at Harvard with Gabriel Kreiman, natural language processing reinforcement learning at Salesforce with Richard Socher, as well as several other projects in industry. Martin also co-founded two startups. Among others, his work has been recognized in the news at Science magazine, MIT News, and Scientific American.
Martin's work has been published at top journals including Neuron and the PNAS as well as leading machine learning venues such as NeurIPS and ICLR where his papers are routinely selected for Oral and Spotlight presentations (0.5% acceptance rate). He has received numerous awards and honors for his research, including the Neuro-Irv and Helga Cooper Open Science Prize, the McGovern fellowship, the Walle Nauta Award for Continuing Dedication in Teaching, the Takeda fellowship in AI Health, the German Federal scholarship, and the MIT Singleton and Shoemaker fellowships. With his startup Integreat, he was a finalist in the Google.org Impact Challenge and won the TUM Social Impact Award, and the Council of Europe's Youth Award.
Education
PhD
Brain and Cognitive Sciences
MIT
2017 - 2022
MSc
Software Engineering
TUM & LMU & UNA
2014 - 2017
BSc
Information Systems
TUM
2011 - 2014
Publications
Selected publications
| Kubilius, J.; Schrimpf, M.; Hong, H.; Majaj, N. J.; Rajalingham, R.; Issa, E. B.; Kar, K.; Bashivan, P.; Prescott-Roy, J.; Schmidt, K.; Nayebi, A.; Bear, D.; Yamins, D. L. K.; DiCarlo, J. J. NeurIPS 2019 (oral) |
Brain-Like Object Recognition with High-Performing Shallow Recurrent ANNs |
| Schrimpf, M.; Kubilius, J.; Lee, M. J.; Ratan Murty, N. A.; Ajemian, R.; DiCarlo, J. J. Neuron 2020 |
Integrative Benchmarking to Advance Neurally Mechanistic Models of Human Intelligence |
| Schrimpf, M.; Blank, I. A.; Tuckute, G.; Kauf, C.; Hosseini, E. A.; Kanwisher, N.; Tenenbaum, J. B.; Fedorenko, E. PNAS 2021 |
The Neural Architecture of Language: Integrative Modeling Converges on Predictive Processing |
| Dapello*, J.; Kar*, K.; Schrimpf, M.; Geary, R.; Ferguson, M.; Cox, D. D.; DiCarlo, J. J. ICLR 2023 (notable top-5%) |
Aligning Model and Macaque Inferior Temporal Cortex Representations Improves Model-to-Human Behavioral Alignment and Adversarial Robustness |
| Bagus, A. M. I. G.; Marques, T.; Sanghavi, S.; DiCarlo, J. J.; Schrimpf, M. NeurIPS SVRHM 2023 |
Primate Inferotemporal Cortex Neurons Generalize Better to Novel Image Distributions Than Analogous Deep Neural Networks Units |
| Tuckute, G.; Sathe, A.; Srikant, S.; Taliaferro, M.; Wang, M.; Schrimpf, M.; Kay, K.; Fedorenko, E. bioRxiv 2023 |
Driving and Suppressing the Human Language Network Using Large Language Models |
| Dapello*, J.; Marques*, T.; Schrimpf, M.; Geiger, F.; Cox, D. D.; DiCarlo, J. J. NeurIPS 2020 (spotlight) |
Simulating a Primary Visual Cortex at the Front of CNNs Improves Robustness to Image Perturbations |
Research
neuroAI
- Building better (= more behaviorally and neurally aligned) models of vision and language. To build these models, we will use neural recordings from non-human primates and humans as well as human behavioral benchmarks, from the Brain-Score platform [see also perspective paper, technical paper, CORnet model, VOneNet model, language paper].
- Integrating multimodal representations: there is a lot of power in shared invariant representations; creating models that go all the way from pixel input to downstream language representations would allow us to potentially harness shared representations between these two domains, and test if we can better map such a model onto brain hierarchy.
- Towards clinical translation: we believe that one of the end goals of these brain-modeling efforts is to apply them and improve people’s lives. This could involve helping blind patients (preliminary work), or people with dyslexia.
Teaching & PhD
Teaching
Computer Science
Communication Systems