The randomized feed-forward neural network is a single hidden layer that enables efficient learning by optimizing only the output weights. ensemble deep framework significantly improves performance of networks. However, framework’s capabilities are limited traditional hyper-parameter selection approaches. Meanwhile, different random architectures, such as existence or lack direct link and mappi...

Among various network compression methods, pruning has developed rapidly due to its superior performance. However, the trivial threshold limits performance of pruning. Most conventional methods are based on well-known hard or soft techniques that rely time-consuming handcrafted tests domain experience. To mitigate these issues, we propose a simple yet effective general method from an optimizati...

Trust and credibility in machine learning models are bolstered by the ability of a model to explain its decisions. While explainability deep is well-known challenge, further challenge clarity explanation itself for relevant stakeholders model. Layer-wise Relevance Propagation (LRP), an established technique developed computer vision, provides intuitive human-readable heat maps input images. We ...

Deep networks are able to learn good representations of unlabelled data via a greedy layer-wise approach to training. One challenge arises in choosing the layer types to use, whether it is an autoencoder, restricted boltzmann machine, with and without sparsity regularization. The layer choice directly affects the type of representations learned. In this paper, we examine sparse autoencoders and...

This paper addresses the following general scenario: A scientist wishes to perform a battery of experiments, each generating a sequential stream of data, to investigate some phenomenon. The scientist would like to control the overall error rate in order to draw statistically-valid conclusions from each experiment, while being as efficient as possible. The between-stream data may differ in distr...

In this paper we establish the existence of path-wise solutions and the uniqueness in law for multidimensional stochastic differential equations driven by a multi-dimensional fractional Brownian motion with Hurst parameter H > 12 .

Image processing and pixel-wise dense prediction have been advanced by harnessing the capabilities of deep learning. One central issue of deep learning is the limited capacity to handle joint upsampling. We present a deep learning building block for joint upsampling, namely guided filtering layer. This layer aims at efficiently generating the highresolution output given the corresponding low-re...

This paper is concerned with the development of new simple 4-noded locking-alleviated smart finite elements for modeling the smart composite beams. The exact solutions for the static responses of the overall smart composite beams are also derived for authenticating the new smart finite elements. The overall smart composite beam is composed of a laminated substrate conventional composite beam, a...

Coherent jets are ubiquitous features of the ocean’s circulation, and their characteristics, such as orientation and energetics, may be influenced by topography. In this study, the authors introduce a large-scale, topographic slope with an arbitrary orientation into quasigeostrophic, doubly periodic, barotropic and baroclinic systems. In both systems, the flow organizes itself into coherent til...

The receptivity of a hypersonic boundary layer to three-dimensional wall perturbations is investigated in this paper by numerical simulations. The work is motivated by Tumin’s [1] theoretical analysis on the receptivity of compressible boundary layers to three-dimensional wall perturbations with the help of the biorthogonal eigenfunction system. Specifically, receptivity processes of a Mach 5.9...