We study the impact of regularization for learning neural networks. Our goal is speeding up training, improving generalization performance, and training compact models that are cost efficient. Our results apply to weight-sharing (e.g. convolutional), sparsity (i.e. pruning), and low-rank constraints among others. We first introduce covering dimension of the constraint set and provide a Rademach...

Deep generative models with multiple hidden layers have been shown to be able to learn meaningful and compact representations of data. In this work we propose deep energy models, which use deep feedforward neural networks to model the energy landscapes that define probabilistic models. We are able to efficiently train all layers of our model simultaneously, allowing the lower layers of the mode...

We give algorithms with provable guarantees that learn a class of deep nets in the generative model view popularized by Hinton and others. Our generative model is an n node multilayer network that has degree at most n for some γ < 1 and each edge has a random edge weight in [−1, 1]. Our algorithm learns almost all networks in this class with polynomial running time. The sample complexity is qua...

The free edges of layered structures require special attention in design processes. Due to the discontinuous material properties, singular stress concentrations occur where layer interfaces meet edge. These bi-material points have be closely examined as possible starting for delamination In present study, we propose an efficient approximate closed-form method analyze interlaminar shear stresses...

The high-order layerwise element models have been used for damaged plates and shells in the presence of singularities such as crack, cutout, and delamination. In this study, the extension of a proposed finite element model has been tested for free vibration analysis of composite laminated systems. For the elements, three-dimensional displacement fields can be captured by layer-by-layer represen...

Abstract Obtaining dense fine ceramics by the binder jetting additive manufacturing process is challenging. A slurry‐based process, such as layerwise slurry deposition (LSD‐print) can enable printing of ceramic parts. This work describes a procedure to develop and qualify suitable ink manufacture silicon carbide green parts LSD‐print. Not only printability but also compatibility with powder bed...