D. Hudson Smith

D. Hudson Smith, John Paul Lineberger, George H. Baker

Many medical ultrasound video recognition tasks involve identifying key anatomical features regardless of when they appear in the video suggesting that modeling such tasks may not benefit from temporal features. Correspondingly, model architectures that exclude temporal features may have better sample efficiency. We propose a novel multi-head attention architecture that incorporates these hypotheses as inductive priors to achieve better sample efficiency on common ultrasound tasks. We compare the performance of our architecture to an efficient 3D CNN video recognition model in two settings: one where we expect not to require temporal features and one where we do. In the former setting, our model outperforms the 3D CNN - especially when we artificially limit the training data. In the latter, the outcome reverses. These results suggest that expressive time-independent models may be more effective than state-of-the-art video recognition models for some common ultrasound tasks in the low-data regime.

D. Hudson Smith, Carl Ehrett, Patrick L. Warren

This paper introduces and tests an unsupervised method for detecting novel coordinated inauthentic information operations (CIOs) in realistic settings. This method uses Bayesian inference to identify groups of accounts that share similar account-level characteristics and target similar narratives. We solve the inferential problem using amortized variational inference, allowing us to efficiently infer group identities for millions of accounts. We validate this method using a set of five CIOs from three countries discussing four topics on Twitter. Our unsupervised approach increases detection power (area under the precision-recall curve) relative to a naive baseline (by a factor of 76 to 580), relative to the use of simple flags or narratives on their own (by a factor of 1.3 to 4.8), and comes quite close to a supervised benchmark. Our method is robust to observing only a small share of messaging on the topic, having only weak markers of inauthenticity, and to the CIO accounts making up a tiny share of messages and accounts on the topic. Although we evaluate the results on Twitter, the method is general enough to be applied in many social-media settings.

D. Hudson Smith, Noah Nisbet, Carl Ehrett et al.

Environmental and individualistic variables affect the rate of human decomposition in complex ways. These effects complicate the estimation of the postmortem interval (PMI) based on observed decomposition characteristics. In this work, we develop a generative probabilistic model for decomposing human remains based on PMI and a wide range of environmental and individualistic variables. This model explicitly represents the effect of each variable, including PMI, on the appearance of each decomposition characteristic, allowing for direct interpretation of model effects and enabling the use of the model for PMI inference and optimal experimental design. In addition, the probabilistic nature of the model allows for the integration of expert knowledge in the form of prior distributions. We fit this model to a diverse set of 2,529 cases from the GeoFOR dataset. We demonstrate that the model accurately predicts 24 decomposition characteristics with an ROC AUC score of 0.85. Using Bayesian inference techniques, we invert the decomposition model to predict PMI as a function of the observed decomposition characteristics and environmental and individualistic variables, producing an R-squared measure of 71%. Finally, we demonstrate how to use the fitted model to design future experiments that maximize the expected amount of new information about the mechanisms of decomposition using the Expected Information Gain formalism.