A5018372746- Neural dynamics and brain function
- Neural Networks and Applications
- Neuroscience and Neural Engineering
- Domain Adaptation and Few-Shot Learning
- Advanced Memory and Neural Computing
- Explainable Artificial Intelligence (XAI)
- Neuroscience and Neuropharmacology Research
- Philosophy and Theoretical Science
- Photoreceptor and optogenetics research
- EEG and Brain-Computer Interfaces
- Reinforcement Learning in Robotics
- Advanced Causal Inference Techniques
- Philosophy and History of Science
- Mental Health and Psychiatry
- Machine Learning and Data Classification
- Robotic Mechanisms and Dynamics
- Adversarial Robustness in Machine Learning
- Receptor Mechanisms and Signaling
- Advanced Bandit Algorithms Research
- Robot Manipulation and Learning
- Cell Image Analysis Techniques
- Functional Brain Connectivity Studies
- Financial Distress and Bankruptcy Prediction
- Scheduling and Optimization Algorithms
- Cell death mechanisms and regulation
St. Jude Children's Research Hospital
2022-2024
University of Pennsylvania
2018-2023
University of Washington
2014-2020
Seattle University
2013-2019
University of Washington Applied Physics Laboratory
2013-2019
California University of Pennsylvania
2018
A critical challenge in neuromorphic computing is to present computationally efficient algorithms of learning. When implementing gradient-based learning, error information must be routed through the network, such that each neuron knows its contribution output, and thus how adjust weight. This known as credit assignment problem. Exactly a solution like backpropagation involves weight sharing, which requires additional bandwidth computations system. Instead, models learning from neuroscience...
The process through which neurons are labeled is a key methodological choice in measuring neuron morphology. However little known about how this may bias measurements. To quantify we compare the extracted morphology of collected from same rodent species, experimental condition, gender distribution, age brain region and putative cell type, but obtained with 19 distinct staining methods. We found strong biases on measured features These were largest related to coverage dendritic tree, e.g....
Prior to receiving visual stimuli, spontaneous, correlated activity called retinal waves drives activity-dependent developmental programs. Early-stage mediated by acetylcholine (ACh) manifest as slow, spreading bursts of action potentials. They are believed be initiated the spontaneous firing Starburst Amacrine Cells (SACs), whose dense, recurrent connectivity then propagates this laterally. Their extended inter-wave intervals and shifting wave boundaries result slow after-hyperpolarization...
Backpropagation is driving today's artificial neural networks (ANNs). However, despite extensive research, it remains unclear if the brain implements this algorithm. Among neuroscientists, reinforcement learning (RL) algorithms are often seen as a realistic alternative: neurons can randomly introduce change, and use unspecific feedback signals to observe their effect on cost thus approximate gradient. convergence rate of such scales poorly with number involved neurons. Here we propose hybrid...
Abstract When a neuron is driven beyond its threshold it spikes, and the fact that does not communicate continuous membrane potential usually seen as computational liability. Here we show this spiking mechanism allows neurons to produce an unbiased estimate of their causal influence, way approximating gradient descent learning. Importantly, neither activity upstream neurons, which act confounders, nor downstream non-linearities bias results. By introducing local discontinuity with respect...
Designing brain-computer interfaces (BCIs) that can be used in conjunction with ongoing motor behavior requires an understanding of how neural activity co-opted for brain control interacts existing circuits. For example, BCIs may to regain lost function after stroke. This controlling unaffected limbs is dissociated from the BCI. In this study we investigated primary cortex accomplishes simultaneous BCI and a task explicitly required both activities driven same region (i.e. dual-control...
Neuroscientists often describe neural activity as a representation of something, or claim to have found evidence for representation. But what do these statements mean? The reasons call some and the assumptions that come with this term are not generally made clear from its common uses in neuroscience. Representation is central concept philosophy mind, rich history going back ancient period. In order clarify usage neuroscience, here we advance link between connotations across disciplines. We...
When a neuron is driven beyond its threshold, it spikes. The fact that does not communicate continuous membrane potential usually seen as computational liability. Here we show this spiking mechanism allows neurons to produce an unbiased estimate of their causal influence, and way approximating gradient descent-based learning. Importantly, neither activity upstream neurons, which act confounders, nor downstream non-linearities bias the results. We how enables solve estimation problems local...
SUMMARY To guide behavior, brain regions such as the orbitofrontal cortex (OFC) retain complex information about current tasks and expected outcomes in cellular representations referred to cognitive maps. When actions produce undesirable results, OFC maps must update promote behavioral change. Here, we show that this remapping is driven by locus coeruleus (LC), a small brainstem nucleus contains most of brain’s norepinephrine (NE)-releasing neurons. In task tests flexibility rodents, LC-NE...
ABSTRACT The combination of fluorescent probes with time-lapse microscopy allows for the visualization entire neuronal activity small animals, such as worms or cnidarians, over a long period time. However, large deformations animal combined natural intermittency make robust automated tracking firing neurons challenging. Here we present an hybrid approach where (i) subset very bright is used moving reference points (fiducials) to estimate elastic deformation animal; (ii) frame-by-frame...
Figure 1 A: Simulated voltage dynamics of model.SACs have Morris-Lecar dynamics, with local coupling occurring via diffusion extracellular ACh, and a noisy channel to induce spontaneous firing.B: Traveling wave speed C as function threshold parameter A refractory timescale ε, computed through numerical continuation.C: Blue points indicate frequency wave-sizes from 5000s simulation.Green line indicates fitted power-law distribution exponent -1.23, coefficient determination r 2 = 0.98.
ethome supports machine learning of animal behavior.It interprets pose-tracking files and behavior annotations to create feature tables, train classifiers, interpolate pose tracking data other common analysis tasks.It features:
Designing brain-computer interfaces (BCIs) that can be used in conjunction with ongoing motor behavior requires an understanding of how neural activity co-opted for brain control interacts existing circuits. For example, BCIs may to regain lost function after stroke. This controlling unaffected limbs is dissociated from the BCI. In this study we investigated primary cortex accomplishes simultaneous BCI and a task explicitly required both activities driven same region (i.e. dual-control...
In good old-fashioned artificial intelligence (GOFAI), humans specified systems that solved problems. Much of the recent progress in AI has come from replacing human insights by learning. However, learning itself is still usually built -- specifically choice parameter updates should follow gradient a cost function. Yet, analogy with GOFAI, there no reason to believe are particularly at defining such systems: we may expect be better if learn it. Recent research machine started realize...
The brain performs unsupervised learning and (perhaps) simultaneous supervised learning. This raises the question as to whether a hybrid of methods will produce better Inspired by rich space Hebbian rules, we set out directly learn rule on local information that best augments signal. We present Hebbian-augmented training algorithm (HAT) for combining gradient-based with an pre-synpatic activity, post-synaptic activities, current weights. test HAT's effect simple problem (Fashion-MNIST) find...