A5047856807- Neural dynamics and brain function
- Visual perception and processing mechanisms
- Cell Image Analysis Techniques
- Face Recognition and Perception
- Advanced Memory and Neural Computing
- Advanced Fluorescence Microscopy Techniques
- Scientific Computing and Data Management
- Neuroscience and Neuropharmacology Research
- Neural Networks and Applications
- EEG and Brain-Computer Interfaces
- Functional Brain Connectivity Studies
- CCD and CMOS Imaging Sensors
- Research Data Management Practices
- stochastic dynamics and bifurcation
- Adversarial Robustness in Machine Learning
- Visual Attention and Saliency Detection
- Memory and Neural Mechanisms
- Advanced Software Engineering Methodologies
- Neurology and Historical Studies
- Neuroscience, Education and Cognitive Function
- Cognitive Science and Mapping
- Categorization, perception, and language
- Color perception and design
- Distributed and Parallel Computing Systems
- Data Analysis with R
University of Washington
2022-2024
Seattle University
2023
DataJoint NEURO (United States)
2019-2023
Centre National de la Recherche Scientifique
2022
CEA Paris-Saclay
2022
Cognitive Neuroimaging Lab
2022
Inserm
2022
Université Paris-Saclay
2022
Commissariat à l'Énergie Atomique et aux Énergies Alternatives
2022
Boston University
2022
Despite great efforts over several decades, our best models of primary visual cortex (V1) still predict spiking activity quite poorly when probed with natural stimuli, highlighting limited understanding the nonlinear computations in V1. Recently, two approaches based on deep learning have emerged for modeling these computations: transfer from artificial neural networks trained object recognition and data-driven convolutional network end-to-end large populations neurons. Here, we test ability...
Abstract The rise of big data in modern research poses serious challenges for management: Large and intricate datasets from diverse instrumentation must be precisely aligned, annotated, processed a variety ways to extract new insights. While high levels integrity are expected, teams have backgrounds, geographically dispersed, rarely possess primary interest science. Here we describe DataJoint, an open-source toolbox designed manipulating processing scientific under the relational model....
Understanding the relationship between circuit connectivity and function is crucial for uncovering how brain implements computation. In mouse primary visual cortex (V1), excitatory neurons with similar response properties are more likely to be synaptically connected, but previous studies have been limited within V1, leaving much unknown about broader rules. this study, we leverage millimeter-scale MICrONS dataset analyze synaptic functional of individual across cortical layers areas. Our...
Abstract To understand the brain we must relate neurons’ functional responses to circuit architecture that shapes them. Here, present a large connectomics dataset with dense calcium imaging of millimeter scale volume. We recorded activity from approximately 75,000 neurons in primary visual cortex (VISp) and three higher areas (VISrl, VISal VISlm) an awake mouse viewing natural movies synthetic stimuli. The data were co-registered volumetric electron microscopy (EM) reconstruction containing...
Divisive normalization (DN) is a prominent computational building block in the brain that has been proposed as canonical cortical operation. Numerous experimental studies have verified its importance for capturing nonlinear neural response properties to simple, artificial stimuli, and suggest DN also an important component processing natural stimuli. However, we lack quantitative models of are directly informed by measurements spiking responses applicable arbitrary Here, propose model input...
Abstract Despite great efforts over several decades, our best models of primary visual cortex (V1) still predict spiking activity quite poorly when probed with natural stimuli, highlighting limited understanding the nonlinear computations in V1. Recently, two approaches based on deep learning have been successfully applied to neural data: On one hand, transfer from networks trained object recognition worked remarkably well for predicting responses higher areas primate ventral stream, but has...
Abstract Despite variations in appearance we robustly recognize objects. Neuronal populations responding to objects presented under varying conditions form object manifolds and hierarchically organized visual areas untangle pixel intensities into linearly decodable representations. However, the associated changes geometry of along cortex remain unknown. Using home cage training showed that mice are capable invariant recognition. We simultaneously recorded responses thousands neurons measure...
We are now in the era of millimeter-scale electron microscopy (EM) volumes collected at nanometer resolution (Shapson-Coe et al., 2021; Consortium 2021). Dense reconstruction cellular compartments these EM has been enabled by recent advances Machine Learning (ML) (Lee 2017; Wu Lu Macrina Automated segmentation methods produce exceptionally accurate reconstructions cells, but post-hoc proofreading is still required to generate large connectomes free merge and split errors. The elaborate 3-D...
In primates and most carnivores, neurons in primary visual cortex are spatially organized by their functional properties. For example, with similar orientation preferences grouped together iso-orientation domains that smoothly vary over the cortical sheet. rodents, on other hand, different thought to be intermingled, a feature which has been termed “salt-and-pepper” organization. The apparent absence of any systematic structure tuning considered defining rodent system for more than decade,...
Abstract To better understand the representations in visual cortex, we need to generate predictions of neural activity awake animals presented with their ecological input: natural video. Despite recent advances models for static images, predicting responses video are scarce and standard linear-nonlinear perform poorly. We developed a new deep recurrent network architecture that predicts inferred spiking thousands mouse V1 neurons simulta-neously recorded two-photon microscopy, while...
Abstract Sensory data about most natural task-relevant variables are entangled with task-irrelevant nuisance variables. The neurons that encode these relevant signals typically constitute a nonlinear population code. Here we present theoretical framework for quantifying how the brain uses or decodes its information. Our theory obeys fundamental mathematical limitations on information content inherited from sensory periphery, describing redundant codes when there many more cortical than...
A bstract Deep neural networks (DNN) have set new standards at predicting responses of populations to visual input. Most such DNNs consist a convolutional network (core) shared across all neurons which learns representation computation in cortex and neuron-specific readout that linearly combines the relevant features this representation. The goal paper is test whether indeed generally characteristic for cortex, i.e. gener-alizes between animals species, what factors contribute obtaining...
Abstract Effective data management is a major challenge for neuroscience labs, and even greater collaborative projects. In the International Brain laboratory (IBL), ten experimental labs spanning 7 geographically distributed sites measure neural activity across brains of mice making perceptual decisions. Here, we report novel, modular architecture that allows users to contribute, access, analyze this collaboration. Users contribute using web-based electronic lab notebook (Alyx), which...
The rise of large scientific collaborations in neuroscience requires systematic, scalable, and reliable data management. How this is best done practice remains an open question. To address this, we conducted a science survey among currently active U19 grants, funded through the NIH’s BRAIN Initiative. was answered by both liaisons Principal Investigators, speaking for ∼500 researchers across 21 nation-wide collaborations. We describe tools, technologies, methods use, identify several...
The relational data model offers unrivaled rigor and precision in defining structure querying complex data. Yet the use of databases scientific pipelines is limited due to their perceived unwieldiness. We propose a simplified conceptually refined named DataJoint. includes language for schema definition, queries, diagramming notation visualizing entities relationships among them. adheres principle entity normalization, which requires that all -- both stored derived must be represented by...
Much of our knowledge about sensory processing in the brain is based on quasi-linear models and stimuli that optimally drive them. However, information nonlinear, even primary areas, optimizing input difficult due to high-dimensional space. We developed inception loops, a closed-loop experimental paradigm combines vivo recordings with silico nonlinear response modeling identify Most Exciting Images (MEIs) for neurons mouse V1. When presented back brain, MEIs indeed drove their target cells...