Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore a novel topological analysis method for understanding neural information processing in this 49-minute workshop lecture from Harvard CMSA's Geometry of Machine Learning series. Discover how the manifold hypothesis applies to neurological systems, where information is represented by lower-dimensional manifolds embedded in higher-dimensional spaces. Learn about the experimental methodology involving micro-electrode array data collection from human and mouse organoids, followed by connectivity extraction using optimized pairwise spike-timing correlations that account for synaptic delays. Examine the network topology identification process for emergent structures and understand the comparison framework using constrained randomization and bootstrapping models. Analyze persistence histograms of topological features that demonstrate how original datasets consistently exceed null distribution variability, indicating significant correlation patterns rather than random fluctuations. Investigate network resiliency findings showing that random 10% node removal maintains significant H1 homology group features (2-dimensional voids), while targeted H1 node removal causes rapid topological collapse, revealing the fragility of brain organoid network cycles. Gain insights into this complementary framework that enhances standard methods for understanding complex neural system information processing through topological analysis applications.
