Learning Optimal Bayesian Networks from Unstructured Data – The objective of this work is to develop a novel method to jointly explore and analyze multiple real world datasets to develop a novel generalization in which data is expressed in a graph, and an inference graph is created that uses that graph to learn the relationships among the data. The resulting graph, the Graph Ontology (GNT), is used to model these two datasets. Experimental results demonstrate the usefulness and efficiency of the proposed method, demonstrating the use of the Graph Ontology to guide the search for new subsets of latent variables, to which we can access relevant data to understand the data.

We present a novel model for age estimation in supervised learning where the task of age estimation is to estimate a new set of informative features (with respect to a set of relevant age labels on that set) from data collected from a population of aging age groups. We present an efficient algorithm for this task, based on a recent novel method for finding informative features for age estimation. The algorithm is fast, yet robust to the non-linearities of the dataset. We compare the performance of existing age estimation algorithms to existing baselines on four benchmark datasets: CIFAR-10, CIFAR-100, CIFAR-200, and VGG51.

Deep Learning with Nonconvex Priors and Nonconvex Loss Functions

AIS-2: Improving, Optimizing and Estimating Multiplicity Optimization

# Learning Optimal Bayesian Networks from Unstructured Data

Learning to rank with hidden measures

Comparing Deep Neural Networks to Matching Networks for Age EstimationWe present a novel model for age estimation in supervised learning where the task of age estimation is to estimate a new set of informative features (with respect to a set of relevant age labels on that set) from data collected from a population of aging age groups. We present an efficient algorithm for this task, based on a recent novel method for finding informative features for age estimation. The algorithm is fast, yet robust to the non-linearities of the dataset. We compare the performance of existing age estimation algorithms to existing baselines on four benchmark datasets: CIFAR-10, CIFAR-100, CIFAR-200, and VGG51.