A5083541135- Neural dynamics and brain function
- Memory and Neural Mechanisms
- Neural Networks and Applications
- Neuroscience and Neuropharmacology Research
- Advanced Memory and Neural Computing
- Photoreceptor and optogenetics research
- Neuroscience and Neural Engineering
- Gene Regulatory Network Analysis
- Functional Brain Connectivity Studies
- Sleep and Wakefulness Research
- Domain Adaptation and Few-Shot Learning
- Cognitive Science and Mapping
- Animal Vocal Communication and Behavior
- Animal Behavior and Reproduction
- Neuroendocrine regulation and behavior
- Ferroelectric and Negative Capacitance Devices
- Zebrafish Biomedical Research Applications
- Robotics and Sensor-Based Localization
- Advanced Image and Video Retrieval Techniques
- Olfactory and Sensory Function Studies
- Machine Learning and ELM
- stochastic dynamics and bifurcation
- Face and Expression Recognition
- Neural Networks and Reservoir Computing
- Visual perception and processing mechanisms
Massachusetts Institute of Technology
2017-2025
McGovern Institute for Brain Research
2019-2025
Institute of Cognitive and Brain Sciences
2020-2023
Princeton University
2023
Intelligent Machines (Sweden)
2023
Moscow Institute of Thermal Technology
2023
The University of Texas at Austin
2012-2021
Brain (Germany)
2020
IIT@MIT
2019
California Institute of Technology
2008-2009
Grid cells in the rat entorhinal cortex display strikingly regular firing responses to animal's position 2-D space and have been hypothesized form neural substrate for dead-reckoning. However, errors accumulate rapidly when velocity inputs are integrated existing models of grid cell activity. To produce grid-cell-like responses, these would require frequent resets triggered by external sensory cues. Such inadequacies, shared various models, cast doubt on dead-reckoning potential system. Here...
Abstract A key challenge in neuroscience is understanding how neurons hundreds of interconnected brain regions integrate sensory inputs with prior expectations to initiate movements. It has proven difficult meet this when different laboratories apply analyses recordings during behaviours. Here, we report a comprehensive set from 115 mice 11 labs performing decision-making task sensory, motor, and cognitive components, obtained 547 Neuropixels probe insertions covering 267 areas the left...
We characterize the relationship between simultaneously recorded quantities of rodent grid cell firing and position rat. The formalization reveals various properties activity when considered as a neural code for representing updating estimates rat9s location. show that, although spatially periodic response cells appears wasteful, is fully combinatorial in capacity. resulting range unambiguous representation vastly greater than ≈1–10 m periods individual lattices, allowing unique...
We propose a model of songbird learning that focuses on avian brain areas HVC and RA, involved in song production, area LMAN, important for generating variability. Plasticity at --> RA synapses is driven by hypothetical "rules" depending three signals: activation synapses, LMAN reinforcement from an internal critic compares the bird's own with memorized template adult tutor's song. Fluctuating glutamatergic input to generates behavioral variability trial-and-error learning. The plasticity...
Neural noise limits the fidelity of representations in brain. This limitation has been extensively analyzed for sensory coding. However, short-term memory and integrator networks, where accumulates can play an even more prominent role, much less is known about how neural interacts with network parameters to determine accuracy computation. Here we analytically derive stored continuous attractor networks probabilistically spiking neurons will degrade over time through diffusion. By combining...
Abstract Research in Neuroscience, as many scientific disciplines, is undergoing a renaissance based on deep learning. Unique to learning models can be used not only tool but interpreted of the brain. The central claims recent learning-based brain circuits are that they make novel predictions about neural phenomena or shed light fundamental functions being optimized. We show, through case-study grid cells entorhinal-hippocampal circuit, one may get neither. begin by reviewing principles cell...
The neural representations of prior information about the state world are poorly understood. To investigate them, we examined brain-wide Neuropixels recordings and widefield calcium imaging collected by International Brain Laboratory. Mice were trained to indicate location a visual grating stimulus, which appeared on left or right with probability alternating between 0.2 0.8 in blocks variable length. We found that mice estimate this thereby improve their decision accuracy. Furthermore,...
Abstract A cognitive map is a suitably structured representation that enables novel computations using previous experience; for example, planning new route in familiar space 1 . Work mammals has found direct evidence such representations the presence of exogenous sensory inputs both spatial 2,3 and non-spatial domains 4–10 Here we tested foundational postulate original theory 1,11 : maps support endogenous without external input. We recorded from entorhinal cortex monkeys mental navigation...
We present a method of estimating the gradient an objective function with respect to synaptic weights spiking neural network. The works by measuring fluctuations in response dynamic perturbation membrane conductances neurons. It is compatible recurrent networks conductance-based model neurons synapses. can be interpreted as biologically plausible learning rule, if perturbations are generated special class "empiric" synapses driven random spike trains from external source.