Deep learning architectures have demonstrated state-of-the-art performance for object classification and have become ubiquitous in commercial products. These methods are often applied without understanding (a) the difficulty of a classification task given the input data, and (b) how a specific deep learning architecture transforms that data. To answer (a) and (b), we illustrate the utility of a multivariate nonparametric estimator of class separation, the Henze-Penrose (HP) statistic, in the original as well as layer-induced representations. Given an $N$-class problem, our contribution defines the $C(N,2)$ combinations of HP statistics as a sample from a distribution of class-pair separations. This allows us to characterize the distributional change to class separation induced at each layer of the model. Fisher permutation tests are used to detect statistically significant changes within a model. By comparing the HP statistic distributions between layers, one can statistically characterize: layer adaptation during training, the contribution of each layer to the classification task, and the presence or absence of consistency between training and validation data. This is demonstrated for a simple deep neural network using CIFAR10 with random-labels, CIFAR10, and MNIST datasets.

David Aha from the Naval Research Laboratory (NRL) gave a talk titled "Machine Learning in the Context Interest in ML and AI is accelerating, and the establishment of Goal Reasoning and Explainable AI." Guna of the JAIC is the DOD's first step in officially Seetharaman, also from NRL, gave a talk titled "Computing recognizing the importance of this area and Architectures: Post Moore's Law and their commitment to building and maintaining the AI/ML/DL Era." Travis Axtell from the Office of the ML and AI communities within the department. Prior Undersecretary of Defense for Intelligence gave a talk to the JAIC, the NAML workshop was one of only titled "AI Ignition," in which he discussed Project a handful of venues to obtain visibility into ML and Maven, an effort that is providing computer vision AI projects within the US Navy and DOD, and no algorithms for object detection, classification, and other venue has such a large and varied audience in alerts in video and still imagery.