skip to main content

Title: Bayesian vector autoregressive model for multi-subject effective connectivity inference using multi-modal neuroimaging data: Bayesian Multi-Modal VAR Model
Award ID(s):
Author(s) / Creator(s):
; ; ; ; ;
Date Published:
Journal Name:
Human Brain Mapping
Page Range / eLocation ID:
1311 to 1332
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. We propose a new order preserving bilinear framework that exploits low-resolution video for person detection in a multi-modal setting using deep neural networks. In this setting cameras are strategically placed such that less robust sensors, e.g. geophones that monitor seismic activity, are located within the field of views (FOVs) of cameras. The primary challenge is being able to leverage sufficient information from videos where there are less than 40 pixels on targets, while also taking advantage of less discriminative information from other modalities, e.g. seismic. Unlike state-of-the-art methods, our bilinear framework retains spatio-temporal order when computing the vector outer products between pairs of features. Despite the high dimensionality of these outer products, we demonstrate that our order preserving bilinear framework yields better performance than recent orderless bilinear models and alternative fusion methods. Code is available at 
    more » « less
  2. Abstract. Plant wax n-alkane chain length distribution and isotopeshave been studied in modern ecosystems as proxies to reconstruct vegetationand climate of the past. However, most paleo-proxies focus on eitherconcentrations or isotopes, whereas both carry complementary information onthe mixing sources. We propose a multi-source mixing model in a Bayesianframework that evaluates both chain length distributions and isotopessimultaneously. The model consists of priors that include user-definedsource groups and their associated parametric distributions of n-alkaneconcentration and δ13C. The mixing process involves newlydefined mixing fractions such as fractional leaf mass contribution (FLMC)that can be used in vegetation reconstruction. Markov Chain Monte Carlo isused to generate samples from the posterior distribution of these parametersconditioned on both data types. We present three case studies from distinctsettings. The first involves n-C27, n-C29, and n-C31 alkanes in lake surface sediments of Lake Qinghai, China. The model provides more specific interpretations on the n-alkane input from aquatic sources than the conventional Paq proxy. The second involves n-C29, n-C31, and n-C33 alkanes in lake surface sediments in Cameroon, western Africa. Themodel produces mixing fractions of forest C3, savanna C3, andC4 plants, offering additional information on the dominant biomescompared to the traditional two-end-member mixing regime. The third couplesthe vegetation source model to a hydrogen isotope model component, usingbiome-specific apparent fractionation factors (εa) toestimate the δ2H of mean annual precipitation. By leveraging chain length distribution, δ13C, and δ2H data offour n-alkane chains, the model produces estimated precipitation δ2H with relatively small uncertainty limits. The new framework shows promise for interpretation of paleo-data but could be further improved by including processes associated with n-alkane turnover in plants, transport,and integration into sedimentary archives. Future studies on modern plantsand catchment systems will be critical to develop calibration datasets thatadvance the strength and utility of the framework. 
    more » « less