Global Mental Health: Five Areas for Value-Driven Training Innovation.

OBJECTIVE: In the field of global mental health, there is a need for identifying core values and competencies to guide training programs in professional practice as well as in academia. This paper presents the results of interdisciplinary discussions fostered during an annual meeting of the Society for the Study of Psychiatry and Culture to develop recommendations for value-driven innovation in global mental health training. METHODS: Participants (n = 48), who registered for a dedicated workshop on global mental health training advertised in conference proceedings, included both established faculty and current students engaged in learning, practice, and research. They proffered recommendations in five areas of training curriculum: values, competencies, training experiences, resources, and evaluation. RESULTS: Priority values included humility, ethical awareness of power differentials, collaborative action, and "deep accountability" when working in low-resource settings in low- and middle-income countries and high-income countries. Competencies included flexibility and tolerating ambiguity when working across diverse settings, the ability to systematically evaluate personal biases, historical and linguistic proficiency, and evaluation skills across a range of stakeholders. Training experiences included didactics, language training, self-awareness, and supervision in immersive activities related to professional or academic work. Resources included connections with diverse faculty such as social scientists and mentors in addition to medical practitioners, institutional commitment through protected time and funding, and sustainable collaborations with partners in low resource settings. Finally, evaluation skills built upon community-based participatory methods, 360-degree feedback from partners in low-resource settings, and observed structured clinical evaluations (OSCEs) with people of different cultural backgrounds. CONCLUSIONS: Global mental health training, as envisioned in this workshop, exemplifies an ethos of working through power differentials across clinical, professional, and social contexts in order to form longstanding collaborations. If incorporated into the ACGME/ABPN Psychiatry Milestone Project, such recommendations will improve training gained through international experiences as well as the everyday training of mental health professionals, global health practitioners, and social scientists.

Learning from Geometry

Subspaces and manifolds are two powerful models for high dimensional signals. Subspaces model linear correlation and are a good fit to signals generated by physical systems, such as frontal images of human faces and multiple sources impinging at an antenna array. Manifolds model sources that are not linearly correlated, but where signals are determined by a small number of parameters. Examples are images of human faces under different poses or expressions, and handwritten digits with varying styles. However, there will always be some degree of model mismatch between the subspace or manifold model and the true statistics of the source. This dissertation exploits subspace and manifold models as prior information in various signal processing and machine learning tasks.

A near-low-rank Gaussian mixture model measures proximity to a union of linear or affine subspaces. This simple model can effectively capture the signal distribution when each class is near a subspace. This dissertation studies how the pairwise geometry between these subspaces affects classification performance. When model mismatch is vanishingly small, the probability of misclassification is determined by the product of the sines of the principal angles between subspaces. When the model mismatch is more significant, the probability of misclassification is determined by the sum of the squares of the sines of the principal angles. Reliability of classification is derived in terms of the distribution of signal energy across principal vectors. Larger principal angles lead to smaller classification error, motivating a linear transform that optimizes principal angles. This linear transformation, termed TRAIT, also preserves some specific features in each class, being complementary to a recently developed Low Rank Transform (LRT). Moreover, when the model mismatch is more significant, TRAIT shows superior performance compared to LRT.

The manifold model enforces a constraint on the freedom of data variation. Learning features that are robust to data variation is very important, especially when the size of the training set is small. A learning machine with large numbers of parameters, e.g., deep neural network, can well describe a very complicated data distribution. However, it is also more likely to be sensitive to small perturbations of the data, and to suffer from suffer from degraded performance when generalizing to unseen (test) data.

From the perspective of complexity of function classes, such a learning machine has a huge capacity (complexity), which tends to overfit. The manifold model provides us with a way of regularizing the learning machine, so as to reduce the generalization error, therefore mitigate overfiting. Two different overfiting-preventing approaches are proposed, one from the perspective of data variation, the other from capacity/complexity control. In the first approach, the learning machine is encouraged to make decisions that vary smoothly for data points in local neighborhoods on the manifold. In the second approach, a graph adjacency matrix is derived for the manifold, and the learned features are encouraged to be aligned with the principal components of this adjacency matrix. Experimental results on benchmark datasets are demonstrated, showing an obvious advantage of the proposed approaches when the training set is small.

Stochastic optimization makes it possible to track a slowly varying subspace underlying streaming data. By approximating local neighborhoods using affine subspaces, a slowly varying manifold can be efficiently tracked as well, even with corrupted and noisy data. The more the local neighborhoods, the better the approximation, but the higher the computational complexity. A multiscale approximation scheme is proposed, where the local approximating subspaces are organized in a tree structure. Splitting and merging of the tree nodes then allows efficient control of the number of neighbourhoods. Deviation (of each datum) from the learned model is estimated, yielding a series of statistics for anomaly detection. This framework extends the classical {\em changepoint detection} technique, which only works for one dimensional signals. Simulations and experiments highlight the robustness and efficacy of the proposed approach in detecting an abrupt change in an otherwise slowly varying low-dimensional manifold.

We Are the Ones We Have Been Waiting for: Pan-African Consciousness Raising and Organizing in the United States and Venezuela

We Are the Ones We Have Been Waiting for: Pan-African Consciousness Raising and Organizing in the United States and Venezuela, draws on fifteen months of field research accompanying organizers, participating in protests, planning/strategy meetings, state-run programs, academic conferences and everyday life in these two countries. Through comparative examination of the processes by which African Diaspora youth become radically politicized, this work deconstructs tendencies to deify political s/heroes of eras past by historicizing their ascent to political acclaim and centering the narratives of present youth leading movements for Black/African liberation across the Diaspora. I employ Manuel Callahan’s description of “encuentros”, “the disruption of despotic democracy and related white middle-class hegemony through the reconstruction of the collective subject”; “dialogue, insurgent learning, and convivial research that allows for a collective analysis and vision to emerge while affirming local struggles” to theorize the moments of encounter, specifically, the moments (in which) Black/African youth find themselves becoming politically radicalized and by what. I examine the ways in which Black/African youth organizing differs when responding to their perpetual victimization by neoliberal, genocidal state-politics in the US, and a Venezuelan state that has charged itself with the responsibility of radically improving the quality of life of all its citizens. Through comparative analysis, I suggest the vertical structures of “representative democracy” dominating the U.S. political climate remain unyielding to critical analyses of social stratification based on race, gender, and class as articulated by Black youth. Conversely, I contend that present Venezuelan attempts to construct and fortify more horizontal structures of “popular democracy” under what Hugo Chavez termed 21st Century Socialism, have resulted in social fissures, allowing for a more dynamic and hopeful negation between Afro-Venezuelan youth and the state.

A Framework to Support the Sharing and Reuse of Computable Phenotype Definitions Across Health Care Delivery and Clinical Research Applications.

INTRODUCTION: The ability to reproducibly identify clinically equivalent patient populations is critical to the vision of learning health care systems that implement and evaluate evidence-based treatments. The use of common or semantically equivalent phenotype definitions across research and health care use cases will support this aim. Currently, there is no single consolidated repository for computable phenotype definitions, making it difficult to find all definitions that already exist, and also hindering the sharing of definitions between user groups. METHOD: Drawing from our experience in an academic medical center that supports a number of multisite research projects and quality improvement studies, we articulate a framework that will support the sharing of phenotype definitions across research and health care use cases, and highlight gaps and areas that need attention and collaborative solutions. FRAMEWORK: An infrastructure for re-using computable phenotype definitions and sharing experience across health care delivery and clinical research applications includes: access to a collection of existing phenotype definitions, information to evaluate their appropriateness for particular applications, a knowledge base of implementation guidance, supporting tools that are user-friendly and intuitive, and a willingness to use them. NEXT STEPS: We encourage prospective researchers and health administrators to re-use existing EHR-based condition definitions where appropriate and share their results with others to support a national culture of learning health care. There are a number of federally funded resources to support these activities, and research sponsors should encourage their use.

Bayesian Learning with Dependency Structures via Latent Factors, Mixtures, and Copulas

Bayesian methods offer a flexible and convenient probabilistic learning framework to extract interpretable knowledge from complex and structured data. Such methods can characterize dependencies among multiple levels of hidden variables and share statistical strength across heterogeneous sources. In the first part of this dissertation, we develop two dependent variational inference methods for full posterior approximation in non-conjugate Bayesian models through hierarchical mixture- and copula-based variational proposals, respectively. The proposed methods move beyond the widely used factorized approximation to the posterior and provide generic applicability to a broad class of probabilistic models with minimal model-specific derivations. In the second part of this dissertation, we design probabilistic graphical models to accommodate multimodal data, describe dynamical behaviors and account for task heterogeneity. In particular, the sparse latent factor model is able to reveal common low-dimensional structures from high-dimensional data. We demonstrate the effectiveness of the proposed statistical learning methods on both synthetic and real-world data.