Integrated Calendar

<p>In this seminar, two new models will be presented. The first new model is a first-principles description of the propagation of light in a cloud, based on a classical solution to Maxwell's equations rather than radiative transfer theory. The second new model is a fully three-dimensional, time-dependent model of the regions of possible sprite inception in the mesosphere, based on the classical method of images from electrostatics rather than finite differencing in space. The reason why each model is unique, the problems each model can solve, and the kinds of results each model can produce will be discussed.&nbsp;&nbsp;</p>

<p>There are two major classes of theories about the basal ganglia. The first class hypothesizes</p><p>that the basal ganglia are the site where cortical sensorimotor and dopaminergic reward</p><p>information interact to potentiate and select actions. These theories predict that content</p><p>specificity of information emerges from within the basal ganglia. The second class of</p><p>theories posits that information is manipulated within the basal ganglia through processes</p><p>such as dimensionality reduction. These theories are primarily based on the fact that there</p><p>is a large reduction in the number of neurons from the input to the output stages of the basal</p><p>ganglia. These theories posit that there are changes in the coding properties of neurons</p><p>rather than the emergence of content specificity.</p><p>In this talk, I will present a set of studies where we analyzed the eye movement system of</p><p>monkeys to compare single-neuron activity in the basal ganglia with activity in the</p><p>cerebellum and the frontal cortex. We used tasks that manipulated both eye movements</p><p>and expected rewards. We found that rather than coding specific sensorimotor or reward</p><p>parameters, the basal ganglia were unique in how they coded these parameters, both in</p><p>terms of the signal-to-noise ratio of responses and in the variety of their temporal patterns.</p><p>These results strongly suggest that the basal ganglia play a role in manipulating rather than</p><p>generating reward and sensorimotor signals.</p>

<p>As life expectancy increases, age-related diseases are becoming more prevalent. While these conditions are traditionally studied in isolation, mounting evidence points to shared, systemic mechanisms underlying these conditions. Our research highlights the vasculature as &nbsp;a key player in organ homeostasis and repair, and a system shared across all organs—making its dysfunction potential driver of age-related pathologies.</p><p>We demonstrate that manipulating <strong>VEGF signaling</strong> to counteract age-related microvascular rarefaction promotes <strong>comprehensive geroprotection</strong>, preserving organ function and delaying disease onset. Our findings also reveal a link between vascular rarefaction and altered RNA splicing. While hypoxia-driven and age-related changes in alternative RNA splicing have been studied independently, we propose a unifying mechanism that links the two. To explore this further, we also employ patient-derived organoids, which retain their biological age in culture, providing a robust in vitro platform to test anti-aging interventions.</p><p>Our findings support a <strong>vascular theory of aging</strong>, identifying vascular health as a promising target to mitigate age-related diseases and promote healthier aging.</p>

Given two distributions P, Q and an integer t, we analyze two sampling processes. In "random allocation," we first sample an index i uniformly from [t], then draw r_{i} ~ P and r_{j} ~ Q for all other j in [t]. In "Poisson sampling," we independently draw r_{i} ~ 1/t*P + (1-1/t)*Q for each i in [t]. We bound the difference between these processes' output distributions and the baseline of sampling r_{i} ~ Q for all i.

This theoretical result provides key insights for analyzing DP-SGD, a privacy-preserving variant of stochastic gradient descent. While Poisson subsampling has well-understood privacy guarantees, common implementations use element shuffling, which was recently shown to have larger privacy losses in certain regimes. Random allocation offers a middle ground, and we prove its privacy analysis reduces to comparing the distributions described above.

We show that these variants' privacy guarantees are within a constant factor of each other across all parameter regimes and converge asymptotically in t. Our proof has two key components: decomposing Poisson sampling into a mixture of random allocation processes, and showing that random allocation can be viewed as a modified Poisson process where sampling probabilities depend on previous outputs.

Joint work with Vitaly Feldman