We introduce Schroedinger Eigenmaps (SE), a new semi-supervised manifold learning and recovery technique. This method is based on an implementation of graph operators with appropriately constructed barrier potentials as carriers labeled information. use our approach for the analysis standard biomedical datasets multispectral retinal images.

This paper considers attacks against machine learning algorithms used in remote sensing applications. The domain presents a suite of challenges that are not fully addressed by current research focused on natural image data. In this we present new study adversarial examples the context satellite classification problems. Using recently curated data set and associated classifier, provide preliminary analysis settings where targeted classifier is permitted multiple observations same location...

Schroedinger Eigenmaps (SE) has recently emerged as a powerful graph-based technique for semi-supervised manifold learning and recovery. By extending the Laplacian of graph constructed from hyperspectral imagery to incorporate barrier or cluster potentials, SE enables machine techniques that employ expert/labeled information provided at subset pixels. In this paper, we show how different types nondiagonal potentials can be used within framework in way allows integration spatial spectral...

The problems of neglecting spatial features in hyperspectral imagery (HSI) and the high complexity Local Tangent Space Alignment (LTSA) still exist nonlinear dimensionality reduction with LTSA for classification. Therefore, this paper proposes an innovative ENH-LTSA (Enhanced-Local Alignment) method to solve two problems. First, random projection is introduced preliminarily reduce dimension HSI data. It aims improve speed neighbor searching local tangent space construction. Then, new...

ABSTRACT Conformability to tissues and adequate mechanical strength are clinically useful properties of resorbable biomaterials used in soft tissue repair. Microbially derived cellulose is attractive as a high strength, highly conformable, biocompatible material for repair, but not naturally resorbable. Here we show that controlled oxidation microbial sheets have been pre‐irradiated with γ‐radiation results fully conformable membrane can be rapidly rehydrated aqueous fluids. In vitro studies...

Abstract This paper discusses the results of a field experiment conducted at Savannah River National Laboratory to test performance several algorithms for localization radioactive materials. In this multirobot system, both an unmanned aerial vehicle, custom hexacopter, and ground vehicle (UGV), ClearPath Jackal, equipped with γ ‐ray spectrometers, were used collect data from two source configurations. Both Fourier scattering transform Laplacian eigenmap detection tested on collected sets....

Data Poisoning attacks modify training data to maliciously control a model trained on such data. In this work, we focus targeted poisoning which cause reclassification of an unmodified test image and as breach integrity. We consider particularly malicious attack that is both "from scratch" "clean label", meaning analyze successfully works against new, randomly initialized models, nearly imperceptible humans, all while perturbing only small fraction the Previous deep neural networks in...

We propose a novel deep learning method which combines classical regularization with data augmentation for estimating myelin water fraction (MWF) in the brain via biexponential analysis. Our aim is to design an accurate technique analysis of signals arising magnetic resonance relaxometry. In particular, we study model, one signal models used MWF estimation. greatly extend our previous work on \emph{input layer (ILR)} several ways. now incorporate optimal parameter selection dedicated neural...

We present Neumann eigenmaps (NeuMaps), a novel approach for enhancing the standard diffusion map embedding using landmarks, i.e distinguished samples within dataset. By interpreting these landmarks as subgraph of larger data graph, NeuMaps are obtained via eigendecomposition renormalized Laplacian. show that offer two key advantages: (1) they provide computationally efficient accurately recovers distance associated with reflecting random walk on subgraph, and (2) naturally incorporate...

We propose a new algorithm to incorporate class conditional information into the critic of GANs via multi-class generalization commonly used Hinge loss that is compatible with both supervised and semi-supervised settings. study compromise between training state art generator an accurate classifier simultaneously, way use our measure degree which are conditional. show trade-off generator-critic pair respecting conditioning inputs generating highest quality images. With multi-hinge...

Abstract There has been growing interest in applying compressed sensing (CS) theory and practice to reconstruct 3D volumes at the nanoscale from electron tomography datasets of inorganic materials, based on known sparsity structure interest. Here we explore application CS for visualizing biological specimens tomographic tilt series acquired scanning transmission microscope (STEM). CS-ET reconstructions match or outperform commonly used alternative methods full undersampled tomogram recovery,...

As new remote sensing modalities emerge, it becomes increasingly important to nd more suitable algorithms for fusion and integration of dierent data types the purposes target/anomaly detection classication. Typical techniques that deal with this problem are based on performing detection/classication/segmentation separately in chosen modalities, then integrating resulting outcomes into a complete picture. In paper we provide broad analysis approach, creating fused representations multi- modal...