Eric V. Strobl

Predicting the risk of psychosis onset: advances and prospects.

To conduct a systematic review of the methods and performance characteristics of models developed for predicting the onset of psychosis.

Deep Multiple Kernel Learning

Deep learning methods have predominantly been applied to large artificial neural networks. Despite their state-of-the-art performance, these large networks typically do not generalize well to datasets with limited sample sizes. In this paper, we take a different approach by learning multiple layers of kernels. We combine kernels at each layer and then optimize over an estimate of the support vector machine leave-one-out error rather than the dual objective function. Our experiments on a variety of datasets show that each layer successively increases performance with only a few base kernels.

Impaired Recognition and Regulation of Disgust Is Associated with Distinct but Partially Overlapping Patterns of Decreased Gray Matter Volume in the Ventroanterior Insula

BACKGROUND The ventroanterior insula is implicated in the experience, expression, and recognition of disgust; however, whether this brain region is required for recognizing disgust or regulating disgusting behaviors remains unknown. METHODS We examined the brain correlates of the presence of disgusting behavior and impaired recognition of disgust using voxel-based morphometry in a sample of 305 patients with heterogeneous patterns of neurodegeneration. Permutation-based analyses were used to determine regions of decreased gray matter volume at a significance level p <= .05 corrected for family-wise error across the whole brain and within the insula. RESULTS Patients with behavioral variant frontotemporal dementia and semantic variant primary progressive aphasia were most likely to exhibit disgusting behaviors and were, on average, the most impaired at recognizing disgust in others. Imaging analysis revealed that patients who exhibited disgusting behaviors had significantly less gray matter volume bilaterally in the ventral anterior insula. A region of interest analysis restricted to behavioral variant frontotemporal dementia and semantic variant primary progressive aphasia patients alone confirmed this result. Moreover, impaired recognition of disgust was associated with decreased gray matter volume in the bilateral ventroanterior and ventral middle regions of the insula. There was an area of overlap in the bilateral anterior insula where decreased gray matter volume was associated with both the presence of disgusting behavior and impairments in recognizing disgust. CONCLUSIONS These findings suggest that regulating disgusting behaviors and recognizing disgust in others involve two partially overlapping neural systems within the insula. Moreover, the ventral anterior insula is required for both processes.

BDNF serum concentrations show no relationship with diagnostic group or medication status in neurodegenerative disease

Brain-derived neurotrophic factor (BDNF) is a growth factor implicated in neuronal survival. Studies have reported altered BDNF serum concentrations in patients with Alzheimers disease (AD). However, these studies have been inconsistent. Few studies have investigated BDNF concentrations across multiple neurodegenerative diseases, and no studies have investigated BDNF concentrations in patients with frontotemporal dementia. To examine BDNF concentrations in different neurodegenerative diseases, we measured serum concentrations of BDNF using enzyme-linked immunoassay in subjects with behavioral-variant frontotemporal dementia (bvFTD, n=20), semantic dementia (SemD, n=16), AD (n=34), and mild cognitive impairment (MCI, n=30), as well as healthy older subjects (HS, n=38). BDNF serum concentrations were compared across diagnoses and correlated with cognitive tests and patterns of brain atrophy using voxelbased morphometry. We found small negative correlations between BDNF serum concentrations and some of the cognitive tests assessing learning, information processing speed and cognitive control in complex situations, however, BDNF did not predict disease group membership despite adequate power. These findings suggest that BDNF serum concentration may not be a reliable diagnostic biomarker to distinguish among neurodegenerative diseases.

Fast causal inference with non-random missingness by test-wise deletion

Many real datasets contain values missing not at random (MNAR). In this scenario, investigators often perform list-wise deletion, or delete samples with any missing values, before applying causal discovery algorithms. List-wise deletion is a sound and general strategy when paired with algorithms such as FCI and RFCI, but the deletion procedure also eliminates otherwise good samples that contain only a few missing values. In this report, we show that we can more efficiently utilize the observed values with test-wise deletion while still maintaining algorithmic soundness. Here, test-wise deletion refers to the process of list-wise deleting samples only among the variables required for each conditional independence (CI) test used in constraint-based searches. Test-wise deletion therefore often saves more samples than list-wise deletion for each CI test, especially when we have a sparse underlying graph. Our theoretical results show that test-wise deletion is sound under the justifiable assumption that none of the missingness mechanisms causally affect each other in the underlying causal graph. We also find that FCI and RFCI with test-wise deletion outperform their list-wise deletion and imputation counterparts on average when MNAR holds in both synthetic and real data.

Markov Boundary Discovery with Ridge Regularized Linear Models.

Abstract Ridge regularized linear models (RRLMs), such as ridge regression and the SVM, are a popular group of methods that are used in conjunction with coefficient hypothesis testing to discover explanatory variables with a significant multivariate association to a response. However, many investigators are reluctant to draw causal interpretations of the selected variables due to the incomplete knowledge of the capabilities of RRLMs in causal inference. Under reasonable assumptions, we show that a modified form of RRLMs can get “very close” to identifying a subset of the Markov boundary by providing a worst-case bound on the space of possible solutions. The results hold for any convex loss, even when the underlying functional relationship is nonlinear, and the solution is not unique. Our approach combines ideas in Markov boundary and sufficient dimension reduction theory. Experimental results show that the modified RRLMs are competitive against state-of-the-art algorithms in discovering part of the Markov boundary from gene expression data.