Lab Positions:
Project descriptions

Postdocs: We have two open postdoctoral positions.

1. We are looking for a postdoctoral candidate to develop robotic control frameworks for mapping signals recorded from human motor cortex to desired manipulation tasks. You will integrate a robotic arm and hand with an existing brain-machine interface, develop new control algorithms for the arm, implement them in software, and help conduct experiments with human participants as part of a multisite clinical trial. This is a joint position with Stanford Robotics Lab, Neural Prosthetics Systems Laboratory, and Neural Prosthetics Translational Lab. It is funded by Stanford Neurosciences Institute.

 

2. We are looking for a postdoctoral candidate to study the underlying mechanisms of spatially localized visual attention. Previous studies found that the resulting perceptual improvements are associated with faster and more reliable spiking in individual neurons, and with less correlated activity fluctuations across the population. Is there a single underlying mechanism that could explain these changes? You will address this question by analyzing and modeling multielectrode recordings from V4 and FEF. This is a joint position with Moore Lab. It is funded by Neuroventures, a Bio-X initiative.

 

Contact Prof. Boahen to find out more about either project, or to propose a project of your own.


Grads
: Rotations are available to participate in several on-going Neuroscience and Neuroengineering projects, some of which are listed above and below. Contact Prof. Boahen to find out more, or to propose your own idea. Taking EE304/BIOE313 (offered in Spring) is a good way to get exposed to neuromorphic engineering.

Undergrads: For research experience, contact a lab-member directly to express your interest in working on his or her project. When applying to Stanford for grad school, chose the program that suits your career objectives. Our lab, like many others at Stanford, has a mix of students from various departments (Bioengineering, Neuroscience, EE/CS, etc.).


Neuroscience
Title Description Requirements
Motor Control
collaboration with Oussama Khatib
How does the brain coordinate the dozens of muscles each limb has, and resolve any potential conflicts on the fly? Control theory developed for articulated robots (by Prof Khatib) yields some testable hypotheses. Evaluate them through theoretical analysis, computational modeling, and fMRI studies.
Knowledge of control theory and exposure to neurobiology of movement.
Visual Attention
collaboration with Tirin Moore
Perceptual improvements conferred by attention are associated with more reliable spiking in individual neurons, and with less correlated activity fluctuations across the population. Is there a single mechanism that could explain these changes? Explore this question by analyzing multielectrode recordings (from the Moore Lab), building computational models, and developing testable hypotheses.
Knowledge of signal proecessing and exposure to systems neuroscience.

Neuroengineering
Title Description Requirements
Optimize NEF Networks for Brainstorm
Using NEF to map computation onto spiking neuron networks reduces memory needed to store weights but increases computation and bandwidth needed for weight updates. Evaluate architectural innovations that mitigate both increases in Brainstorm by simulating them with a customized version of Nengo.
Experience with Python and exposure to neural networks.
Design Neuromorph's Place & Route Algorithm
Brainstorm will have hundreds of arrays with thousands of neurons each. Their weights will be stored in local RAM, adjacent to each array. An interconnection network will communicate spikes between arrays. Design an optimization algorithm to assign NEF ensembles to arrays as closely together as possible and pack their decoders, transforms & encoders in local RAM as tightly as possible so as to minimize spike traffic.
Programming expertise (C++ or Python) and knowledge of optimization algorithms.
Make Brainstorm's Computation Temperature-
Invariant
Use dimensionless neural models that capture how transistor mismatch in subthreshold analog circuits depends on temperature to predict changes in tuning curves of Brainstorm's silicon neurons. Your success will enable Neuromorph to derive decoders that map computations onto Brainstorm in a temperature-invariant fashion
Expertise in analog circuits and familiarity with MatLab and/or Phyton.
Prosthesis
collaboration with Krishna Shenoy
Current brain-machine-interfaces dissipate too much power to be implanted inside the brain. Neuromorphic chips could provide an ultra-low power alternative. Explore this by using NEF to map decoders for spike trains recorded from motor cortex and controllers for robotic arms and hands onto spiking neural networks.
Knowledge of control theory (e.g., Kalman Filter) and familiarity with neural networks.
Design a Daughterboard for Neurogrid
Increase the number of weighted synaptic connections that Neurogrid supports and upgrade its computer interface from USB 2.0 to USB 3.0 by designing a new daughterboard. This new board will enable us to implement spiking neural networks that display interesting cognitive visuomotor behaviors using NEF.
Experience with PCB design and familiarity with FPGA programming (e.g., in VHDL).
Design new ways to express semantic relationships
Can we use neuromorphic chips to implement human-like behavio? One strategy for doing is is the semantic pointer architecture. This relies on the assumption that ideas ("boy", "ownership", "ball") can be expressed as vectors and that relationships between them ("the boy owns the ball") are described using an operation called circular convolution. A much more efficient approach is to use permuations to accomplish the same task. Explore the properties and uses of permutation encodings and design neural populations that implement them efficiently. Familiarity with linear algebra and some programming experience.

Learning
Title Description Requirements
Design new ways to express semantic relationships.
How does the brain coordinate the dozens of muscles each limb has, and resolve any potential conflicts on the fly? Control theory developed for articulated robots (by Prof Khatib) yields some testable hypotheses. Evaluate them through theoretical analysis, computational modeling, and fMRI studies.
Knowledge of control theory and exposure to neurobiology of movement.
Visual Attention
collaboration with Tirin Moore
Perceptual improvements conferred by attention are associated with more reliable spiking in individual neurons, and with less correlated activity fluctuations across the population. Is there a single mechanism that could explain these changes? Explore this question by analyzing multielectrode recordings (from the Moore Lab), building computational models, and developing testable hypotheses.
Knowledge of signal proecessing and exposure to systems neuroscience.