A central goal in the cognitive sciences is the development of numerical models for mental representations of object concepts. This paper introduces Variational Interpretable Concept Embeddings (VICE), an approximate Bayesian method for embedding object concepts in a vector space using data collected from humans in a triplet odd-one-out task. VICE uses variational inference to obtain sparse, non-negative representations of object concepts with uncertainty estimates for the embedding values. These estimates are used to automatically select the dimensions that best explain the data. We derive a PAC learning bound for VICE that can be used to estimate generalization performance or determine a sufficient sample size for experimental design. VICE rivals or outperforms its predecessor, SPoSE, at predicting human behavior in the triplet odd-one-out task. Furthermore, VICE's object representations are more reproducible and consistent across random initializations, highlighting the unique advantage of using VICE for deriving interpretable embeddings from human behavior.

Collecting the large datasets needed to train deep neural networks can be very difficult, particularly for the many applications for which sharing and pooling data is complicated by practical, ethical, or legal concerns. However, it may be the case that derivative datasets or predictive models developed within individual sites can be shared and combined with fewer restrictions. Training on distributed data and combining the resulting networks is often viewed as continual learning, but these methods require networks to be trained sequentially. In this paper, we introduce distributed weight consolidation (DWC), a continual learning method to consolidate the weights of separate neural networks, each trained on an independent dataset. We evaluated DWC with a brain segmentation case study, where we consolidated dilated convolutional neural networks trained on independent structural magnetic resonance imaging (sMRI) datasets from different sites. We found that DWC led to increased performance on test sets from the different sites, while maintaining generalization performance for a very large and completely independent multi-site dataset, compared to an ensemble baseline.

The difficulty of multi-class classification generally increases with the number of classes. Using data for a small set of the classes, can we predict how well the classifier scales as the number of classes increases? We propose a framework for studying this question, assuming that classes in both sets are sampled from the same population and that the classifier is based on independently learned scoring functions. Under this framework, we can express the classification accuracy on a set of $k$ classes as the $(k - 1)$st moment of a discriminability function; the discriminability function itself does not depend on $k$. We leverage this result to develop a non-parametric regression estimator for the discriminability function, which can extrapolate accuracy results to larger unobserved sets. We also formalize an alternative approach that extrapolates accuracy separately for each class, and identify tradeoffs between the two methods. We show that both methods can accurately predict classifier performance on label sets up to ten times the size of the original set, both in simulations as well as in realistic face recognition or character recognition tasks.