Institution
Stevens Institute of Technology
Education•Hoboken, New Jersey, United States•
About: Stevens Institute of Technology is a education organization based out in Hoboken, New Jersey, United States. It is known for research contribution in the topics: Computer science & Cognitive radio. The organization has 5440 authors who have published 12684 publications receiving 296875 citations. The organization is also known as: Stevens & Stevens Tech.
Topics: Computer science, Cognitive radio, Communication channel, Wireless network, Artificial neural network
Papers published on a yearly basis
Papers
More filters
••
03 Apr 2020TL;DR: This study investigates the effects of Markov chain Monte Carlo (MCMC) sampling in unsupervised Maximum Likelihood (ML) learning and identifies a variety of ML learning outcomes that depend solely on the implementation of MCMC sampling.
Abstract: This study investigates the effects of Markov chain Monte Carlo (MCMC) sampling in unsupervised Maximum Likelihood (ML) learning. Our attention is restricted to the family of unnormalized probability densities for which the negative log density (or energy function) is a ConvNet. We find that many of the techniques used to stabilize training in previous studies are not necessary. ML learning with a ConvNet potential requires only a few hyper-parameters and no regularization. Using this minimal framework, we identify a variety of ML learning outcomes that depend solely on the implementation of MCMC sampling.On one hand, we show that it is easy to train an energy-based model which can sample realistic images with short-run Langevin. ML can be effective and stable even when MCMC samples have much higher energy than true steady-state samples throughout training. Based on this insight, we introduce an ML method with purely noise-initialized MCMC, high-quality short-run synthesis, and the same budget as ML with informative MCMC initialization such as CD or PCD. Unlike previous models, our energy model can obtain realistic high-diversity samples from a noise signal after training.On the other hand, ConvNet potentials learned with non-convergent MCMC do not have a valid steady-state and cannot be considered approximate unnormalized densities of the training data because long-run MCMC samples differ greatly from observed images. We show that it is much harder to train a ConvNet potential to learn a steady-state over realistic images. To our knowledge, long-run MCMC samples of all previous models lose the realism of short-run samples. With correct tuning of Langevin noise, we train the first ConvNet potentials for which long-run and steady-state MCMC samples are realistic images.
79 citations
••
TL;DR: This paper presents some results regarding the design of reliable networks and solves the problem of synthesizing graphs in a class of p vertex, q edge graphs with the property that any graph in the class has a smaller probability of disconnection than any graph outside of the class.
Abstract: This paper presents some results regarding the design of reliable networks. The problem under consideration involves networks which are undirected graphs having equal and independent edge failure probabilities. The index of reliability is the probability that the network fails (becomes disconnected). For “small” edge failure probabilities and given p and q there exists a class of p vertex, q edge graphs with the property that any graph in the class has a smaller probability of disconnection than any graph outside of the class. We solve the problem of synthesizing graphs in this class.
79 citations
••
TL;DR: Both the morphology and conductivity of Cu2O films are controlled in a facile electrodeposited process by tuning the concentration of surfactants by affecting the reduction rate of Cu(2+) ions and the formation of oxygen vacancies or copper vacancies during the electrodeposition.
Abstract: Both the morphology and conductivity of Cu2O films are controlled in a facile electrodeposition process by tuning the concentration of surfactants. With the increase of the concentration of sodium dodecyl sulfate (SDS) in the plating solution, the average size of Cu2O crystals increases, and the electrical conductivity of Cu2O films changes from n-type to p-type. When the concentrations of SDS are lower than 0.85 mM, the electrodeposited Cu2O films show n-type conductivity because of the formation of oxygen vacancies or copper atoms. When the concentration of SDS is higher than 1.70 mM, the electrodeposited Cu2O films show p-type conductivity owing to the formation of copper vacancies. The concentrations of both the donors and the acceptors increase with the concentration of SDS. The effects of surfactants on the morphology and conductivity of electrodeposited Cu2O films are attributed to the adsorption of SDS molecules on the electrode substrate occupying the deposition sites of Cu(2+) ions and the adsorption of SDS micelles to Cu(2+) ions hindering the diffusion of Cu(2+) ions to the electrode, which affect the reduction rate of Cu(2+) ions and the formation of oxygen vacancies or copper vacancies during the electrodeposition.
79 citations
••
TL;DR: In this paper, the authors used a broadband laser vibrometer (dc-1.5 MHz) as a detector to detect 3D elastic solids (doped glass and Berea sandstone).
Abstract: Reverberant volume time reversal in 3D elastic solids (doped glass and Berea sandstone) using a single channel are presented. In spite of large numbers of mode conversions (compressional to shear wave conversions at the walls), time reversal works extremely well, providing very good spatial and time focusing of elastic waves. Ceramics were bonded to the surface as sources (100–700 kHz); a broadband laser vibrometer (dc—1.5 MHz) was used as detector. Temporal and spatial time-reversal focusing are frequency dependent and depend on the dissipation characteristics of the medium. Doped glass (inverse dissipation Q between 2000 to 3000) shows time-reversed spatial focal resolution at about half of the shear wavelength. The Berea sandstone (Q=50) yields a wider focusing width (a bit more than the shear wavelength) due to its lower Q. Focusing in the doped glass is better because the time-reversal (virtual) array created by wave reflections is larger than in the highly attenuating sandstone. These are the first results reported in granular media, and are a first step toward geophysical and field applications.
79 citations
••
TL;DR: It is demonstrated that utilization of nonpolar polystyrene as a cavity dielectric completely removes spectral diffusion and blinking in individual SWCNTs on the millisecond to multisecond time scale, despite the presence of surfactants.
Abstract: Single-walled carbon nanotubes (SWCNTs) are considered for novel optoelectronic and quantum photonic devices, such as single photon sources, but methods must be developed to enhance the light extraction and spectral purity, while simultaneously preventing multiphoton emission as well as spectral diffusion and blinking in dielectric environments of a cavity. Here we demonstrate that utilization of nonpolar polystyrene as a cavity dielectric completely removes spectral diffusion and blinking in individual SWCNTs on the millisecond to multisecond time scale, despite the presence of surfactants. With these cavity-embedded SWCNT samples, providing a 50-fold enhanced exciton emission into the far field, we have been able to carry out photophysical studies for the first time with nanosecond timing resolution. We uncovered that fast spectral diffusion processes (1–3 ns) remain that make significant contributions to the spectral purity, thereby limiting the use of SWCNTs in quantum optical applications requiring i...
79 citations
Authors
Showing all 5536 results
Name | H-index | Papers | Citations |
---|---|---|---|
Paul M. Thompson | 183 | 2271 | 146736 |
Roger Jones | 138 | 998 | 114061 |
Georgios B. Giannakis | 137 | 1321 | 73517 |
Li-Jun Wan | 113 | 639 | 52128 |
Joel L. Lebowitz | 101 | 754 | 39713 |
David Smith | 100 | 994 | 42271 |
Derong Liu | 77 | 608 | 19399 |
Robert R. Clancy | 77 | 293 | 18882 |
Karl H. Schoenbach | 75 | 494 | 19923 |
Robert M. Gray | 75 | 371 | 39221 |
Jin Yu | 74 | 480 | 32123 |
Sheng Chen | 71 | 688 | 27847 |
Hui Wu | 71 | 347 | 19666 |
Amir H. Gandomi | 67 | 375 | 22192 |
Haibo He | 66 | 482 | 22370 |