Welcome to the MacLean Lab!
A Multi-Disciplinary Approach to NeocortexWhat We Do
We are a group of multi-disciplinary scientists who employ computational and empirical approaches to the study of neocortex. Our research focuses on the mechanisms of neuronal circuit spiking dynamics and how these dynamics control behavior. We employ a blend of large-scale neuronal recordings, quantitative analysis, computational modeling, and machine learning. Instead of specializing in any one region of neocortex, we are interested in identifying commonalities across regions.
Recent projects include: identifying the stiff and sloppy dimensions of synaptic architectures capable of generating spiking matched to mouse cortex, the impact of neocortical features on task training and performance of recurrent spiking neural network models, and the changes in information flow during a reach-to-grasp task as revealed using graph theoretic tools.
Find us on GitHub!
Recent Projects
Stiff and Sloppy Dimensions in Synaptic Architectures
How cortex varies its connections during learning without collapse into excessively high- or low- activity states is poorly understood. Using simulations on neural networks with first-order spiking statistics, matched to firing in murine visual cortex, while varying connectivity parameters, we determined the stiff and sloppy parameters of synaptic architectures across three classes of input (brief, continuous, and cyclical). Algorithmically-generated connectivity parameter values drawn from a large portion of the parameter space reveal that specific combinations of excitatory and inhibitory connectivity are stiff and that all other architectural details are sloppy. That these characteristics emerge consistently from an unbiased generation scheme provides a novel vector of affirmation for the importance of interrelation between excitatory and inhibitory connectivity in establishing and maintaining stable spiking dynamical regimes in neocortex.
Simulating Neocortex with Spiking Neural Networks
Formalizing the distinction between computation-performing and computation-nonperforming neural networks spearheads broader analyses into how cortex organizes itself when learning. Structural and dynamical features provide a system of metrics and methodologies which provide this formalization. Taking advantage of recent advances in biologically-plausible machine learning, we have developed a spiking neural network to serve as a model system capable of completing meanigful computations in which such features can be asserted or allowed to emerge dynamically in the variably controlled presence or absence of combinations of neocortical features of current interest to the neuroscientific community. This toolset provides a novel vector of analysis on the importance and role of specific features and feature-interactions in neural network’s, with particular value in assessing network performance and identifying how evolved characteristics of neoxortex support computation.
Murine Information Flow Dynamics during Reach-to-Grab
Learning to precisely execute dexterous movements drives plasticity in motor cortex, a densely interconnected, multi-layered brain structure. Thus far our understanding of this plasticity, and the subsequent changes in neural activity, comes from studying individual layers separately. Using an optical prism attached to a miniaturized microscope, it’s possible to simultaneously track the activity of many neurons in multiple cortical layers, and to do so in freely moving animal subjects over many days as they learn a new skill. Combining these results with detailed kinematics data, on a scale only made possible by modern computer vision technologies, toolsets from network science, graph theory, and information theory are jointly used to identify and track the interplay between cortical network components over time. This long-term examination of learning-driven plasticity seeks to provide novel insight to functional networks’ dynamic graph topology within and across layers.
Get in Touch!
Researchers welcome. Articles available.