Scientist – Biological constraints of computation
Our mission at the Allen Institute is to advance our understanding of how the brain works in health and disease. Using a team science approach, we strive to discover how the brain implements fundamental computations through the integration of technological innovation, cutting-edge experiments, modeling, and theory.
The defining aspect of the brain is its capacity to produce behavior. However, how behavior arises from the computations implemented by different systems in the brain, what is the set of computations implemented by different systems, and how this set of computations is limited by anatomical constraints remains poorly understood.
We want to use anatomical data (http://connectivity.brain-map.org/) to constrain the space of architectures for the visual system, train models for different etologically relevant tasks, and compare model’s responses with those physiological observed (https://observatory.brain-map.org/visualcoding). One of the salient characteristics of the responses in the mouse visual system is the large variability of responses to the same images, and the combination of visual and motor responses. We seek to explore the space of tasks which produce similar representations to those observed in the mouse brain, with a focus on development of mixed visual/motor representations. The final results of the models trained on etologically relevant tasks can be used as initial states for models learning.
We are seeking a scientist to join the Modeling and Theory team of Stefan Mihalas to help construct biologically constrained task driven models of the mouse cortex. The scientist will help transform biological knowledge of structure into architectural constraints, train models with such over a range of tasks, and compare the results to in vivo physiology measurements. The scientist is expected to have strong contributions at the level of ideas, and help shape future projects. Our team focuses at extracting principles from large data, integrating them in simplified models and using the models to test theories of computation.
Construct architectural constraints for the network from large scale connectivity data
Train models for visual processing with architectural constraints to perform different visual/motor tasks, and compare the results with the physiological recordings
In collaboration with experimentalists contribute to the analysis of the anatomical and physiological data
Contribute scientific ideas based on the modeling results, and help shape future projects
Develop and maintain computational and associated software tools
Maintain clear and accurate communication with supervisor and team members
Publish/present findings in peer-reviewed journals/scientific conferences
Required Experience and Education
PhD degree in computational neuroscience, physics, mathematics, applied mathematics or related field.
Scientist I: 0-2 years of post-doctoral experience
Strong background in scientific computing and data analysis
Proven ability to code using Python and experience using PyTorch or TensorFLow
Strong written and verbal communication skills.
Preferred Education and Experience
Experience in computational neuroscience
Ability to meet aggressive timelines and deliverables in a collaborative environment.
Position Type/Expected Hours of Work
It is the policy of the Allen Institute to provide equal employment opportunity (EEO) to all persons regardless of age, color, national origin, citizenship status, physical or mental disability, race, religion, creed, gender, sex, sexual orientation, gender identity and/or expression, genetic information, marital status, status with regard to public assistance, veteran status, or any other characteristic protected by federal, state or local law. In addition, the Allen Institute will provide reasonable accommodations for qualified individuals with disabilities.