Internal Seminar: Doya Unit and Economo Unit

Join us for April's second Internal Seminar Series on Apr. 17, from 17:00 to 18:00 in B250. This month's second seminar features the Neural Computation Unit (Kenji Doya) and Biodiversity and Biocomplexity Unit (Evan Economo).

Neural Computation Unit (Kenji Doya)

Speaker : Yu Shimizu

Title : Machine Learning and Mental Disease - Automated Diagnosis and Subtype Identification in Major Depression

Abstract : Advances in imaging methods together with drastic enhancement of computational power have enabled detailed measurement of brain activation, consequently sparking a search for efficient and objective diagnosis methods for mental disease. This search is a search for sophisticated data mining (machine learning) algorithms, which can "learn" to distinguish data of healthy subjects from that of diseased with high accuracy. Ideally, they should also provide insight into the specific characteristics of the data driving the distinction. As will be explained in this talk, not all algorithms provide such information. Using our own project on major depression as an example, we will give an idea on the possibilities and challenges of big data analysis using machine learning, its implication on the diagnosis of mental diseases and its hopes for prevention and treatment programs.

Biodiversity and Biocomplexity Unit (Evan Economo)

Speaker : Evan Economo

Title : Global Patterns in the Evolution of Biodiversity

Abstract : Resolving global biodiversity patterns and understanding their evolutionary and ecological drivers remain key challenges for basic and applied biology. Recently, large-scale data syntheses and analyses have opened a broad spatial and temporal view of biodiversity patterns and dynamics, but this progress has been mostly limited to vertebrates and plants. Here, we compile and analyze a global distribution database for all 15,000+ described ant species, consisting of over 1.6 million literature and specimen records. Using these data, we identify the main gradients and hotspots of species richness around the globe. To investigate the evolutionary history underlying spatial richness patterns, we reconstruct time-calibrated phylogenies covering all ant diversity and perform Bayesian analysis of diversification rate variation across the tree. We find complex patterns of diversification rate variation but little effect of latitude despite a steep latitudinal gradient in species richness. We also found that overall temperate taxa are highly phylogenetically clustered and nested within tropical clades, but that transitions out of the tropical habitats only partially coincide with climatic cooling periods.