K. Miller

K. Miller
Are you K. Miller?

Claim your profile, edit publications, add additional information:

Contact Details

Name
K. Miller
Affiliation
Location

Pubs By Year

Pub Categories

 
Instrumentation and Methods for Astrophysics (9)
 
Physics - Materials Science (7)
 
Computer Science - Networking and Internet Architecture (6)
 
Computer Science - Learning (5)
 
Astrophysics (4)
 
Statistics - Machine Learning (4)
 
Quantitative Biology - Neurons and Cognition (4)
 
Physics - Instrumentation and Detectors (3)
 
Physics - Atomic Physics (3)
 
Computer Science - Multimedia (3)
 
Physics - Strongly Correlated Electrons (2)
 
Physics - Optics (2)
 
Physics - Other (2)
 
Statistics - Applications (2)
 
Computer Science - Neural and Evolutionary Computing (1)
 
Quantitative Biology - Quantitative Methods (1)
 
Mathematics - Numerical Analysis (1)
 
Computer Science - Artificial Intelligence (1)
 
Computer Science - Architecture (1)
 
Mathematics - Probability (1)
 
Computer Science - Distributed; Parallel; and Cluster Computing (1)
 
Physics - Biological Physics (1)
 
High Energy Physics - Experiment (1)
 
Cosmology and Nongalactic Astrophysics (1)
 
Earth and Planetary Astrophysics (1)
 
Nuclear Experiment (1)
 
Computer Science - Computers and Society (1)
 
Astrophysics of Galaxies (1)
 
Computer Science - Databases (1)
 
Mathematics - History and Overview (1)

Publications Authored By K. Miller

The success of deep neural networks has inspired many to wonder whether other learners could benefit from deep, layered architectures. We present a general framework called forward thinking for deep learning that generalizes the architectural flexibility and sophistication of deep neural networks while also allowing for (i) different types of learning functions in the network, other than neurons, and (ii) the ability to adaptively deepen the network as needed to improve results. This is done by training one layer at a time, and once a layer is trained, the input data are mapped forward through the layer to create a new learning problem. Read More

We study the problem of interactively learning a binary classifier using noisy labeling and pairwise comparison oracles, where the comparison oracle answers which one in the given two instances is more likely to be positive. Learning from such oracles has multiple applications where obtaining direct labels is harder but pairwise comparisons are easier, and the algorithm can leverage both types of oracles. In this paper, we attempt to characterize how the access to an easier comparison oracle helps in improving the label and total query complexity. Read More

Many architects believe that major improvements in cost-energy-performance must now come from domain-specific hardware. This paper evaluates a custom ASIC---called a Tensor Processing Unit (TPU)---deployed in datacenters since 2015 that accelerates the inference phase of neural networks (NN). The heart of the TPU is a 65,536 8-bit MAC matrix multiply unit that offers a peak throughput of 92 TeraOps/second (TOPS) and a large (28 MiB) software-managed on-chip memory. Read More

Direct imaging of exoplanets requires establishing and maintaining a high contrast dark hole (DH) within the science image to a high degree of precision. Current approaches aimed at establishing a DH, such as electric field conjugation (EFC), have been demonstrated in the lab and have proven capable of high contrast DH generation. The same approaches have been considered for the maintenance of the DH as well. Read More

Purpose: To investigate the effect of realistic microstructural geometry on the susceptibility-weighted magnetic resonance (MR) signal in white matter (WM), with application to demyelination. Methods: Previous work has modeled susceptibility-weighted signals under the assumption that axons are cylindrical. In this work, we explore the implications of this assumption by considering the effect of more realistic geometries. Read More

Infrared (IR) blocking filters are crucial for controlling the radiative loading on cryogenic systems and for optimizing the sensitivity of bolometric detectors in the far-IR. We present a new IR filter approach based on a combination of patterned frequency selective structures on silicon and a thin (50 $\mu \textrm{m}$ thick) absorptive composite based on powdered reststrahlen absorbing materials. For a 300 K blackbody, this combination reflects $\sim$50\% of the incoming light and blocks \textgreater 99. Read More

Refractive optical elements are widely used in millimeter and sub-millimeter astronomical telescopes. High resistivity silicon is an excellent material for dielectric lenses given its low loss-tangent, high thermal conductivity and high index of refraction. The high index of refraction of silicon causes a large Fresnel reflectance at the vacuum-silicon interface (up to 30%), which can be reduced with an anti-reflection (AR) coating. Read More

We report on the scrambling performance and focal-ratio-degradation (FRD) of various octagonal and rectangular fibers considered for MAROON-X. Our measurements demonstrate the detrimental effect of thin claddings on the FRD of octagonal and rectangular fibers and that stress induced at the connectors can further increase the FRD. We find that fibers with a thick, round cladding show low FRD. Read More

We investigate the utility of coded aperture (CA) for roadside radiation threat detection applications. With coded aperture, information in the form of photon quantity is traded for directional information. Whether and in what scenarios this trade-off is beneficial is the focus of this study. Read More

TCPTuner is a TCP (transmission control protocol) congestion control kernel module and GUI (graphical user interface) for Linux that allows real-time modification of the congestion control parameters of TCP CUBIC, the current default algorithm in Linux. Specifically, the tool provides access to alpha, the rate at which a sender's congestion window grows; beta, the multiplicative factor to decrease the congestion window on a loss event; as well as CUBIC's fast convergence and tcp friendliness parameters. Additionally, the interface provides access to ip-route parameters for the minimum retransmission time and initial congestion window size. Read More

We have incorporated our experimentally derived thermal rate coefficients for C + H$_3^+$ forming CH$^+$ and CH$_2^+$ into a commonly used astrochemical model. We find that the Arrhenius-Kooij equation typically used in chemical models does not accurately fit our data and use instead a more versatile fitting formula. At a temperature of 10 K and a density of 10$^4$ cm$^{-3}$, we find no significant differences in the predicted chemical abundances, but at higher temperatures of 50, 100, and 300 K we find up to factor of 2 changes. Read More

The experimental investigation of a broadband far-infrared meta-material absorber is described. The observed absorptance is $>\,0.95$ from ${\rm 1-20\,THz}$ (${\rm 300-15\,\mu m}$) over a temperature range spanning ${\rm 5-300\,K}$. Read More

Silicon oxide thin films play an important role in the realization of optical coatings and high-performance electrical circuits. Estimates of the dielectric function in the far- and mid-infrared regime are derived from the observed transmittance spectrum for a commonly employed low-stress silicon oxide formulation. The experimental, modeling, and numerical methods used to extract the dielectric function are presented. Read More

Recently, HTTP-based adaptive streaming has become the de facto standard for video streaming over the Internet. It allows clients to dynamically adapt media characteristics to network conditions in order to ensure a high quality of experience, that is, minimize playback interruptions, while maximizing video quality at a reasonable level of quality changes. In the case of live streaming, this task becomes particularly challenging due to the latency constraints. Read More

For several years the pervasive belief that the Super Bowl is the single biggest day for human trafficking in the United States each year has been perpetuated in popular press despite a lack of evidentiary support. The practice of relying on hearsay and popular belief for decision-making may result in misappropriation of resources in anti-trafficking efforts. We propose a data-driven approach to analyzing sex trafficking, especially as it is carried on during--and perhaps in response to--large public events such as the Super Bowl. Read More

Internet networks are handling increasing volume of traffic than ever before. This data is mainly associated to sensitive, distributed, and multimedia applications. In the past years, much attention has been paid to the way network infrastructure must be designed and developed in order to handle the challenges of delivering high quality services for applications such as VoIP and streaming video. Read More

The COHERENT collaboration's primary objective is to measure coherent elastic neutrino-nucleus scattering (CEvNS) using the unique, high-quality source of tens-of-MeV neutrinos provided by the Spallation Neutron Source (SNS) at Oak Ridge National Laboratory (ORNL). In spite of its large cross section, the CEvNS process has never been observed, due to tiny energies of the resulting nuclear recoils which are out of reach for standard neutrino detectors. The measurement of CEvNS has now become feasible, thanks to the development of ultra-sensitive technology for rare decay and weakly-interacting massive particle (dark matter) searches. Read More

Human trafficking is a challenging law enforcement problem, with a large amount of such activity taking place online. Given the large, heterogeneous and noisy structure of this data, the problem becomes even more challenging. In this paper we propose and entity resolution pipeline using a notion of proxy labels, in order to extract clusters from this data with prior history of human trafficking activity. Read More

One significant challenge to scaling entity resolution algorithms to massive datasets is understanding how performance changes after moving beyond the realm of small, manually labeled reference datasets. Unlike traditional machine learning tasks, when an entity resolution algorithm performs well on small hold-out datasets, there is no guarantee this performance holds on larger hold-out datasets. We prove simple bounding properties between the performance of a match function on a small validation set and the performance of a pairwise entity resolution algorithm on arbitrarily sized datasets. Read More

We demonstrate the use of high power diode laser stacks to photodetach fast hydrogen and carbon anions and produce ground term neutral atomic beams. We achieve photodetachment efficiencies of $\sim$7.4\% for H$^-$ at a beam energy of 10\,keV and $\sim$3. Read More

We have measured the reaction of O + H3+ forming OH+ and H2O+. This is one of the key gas-phase astrochemical processes initiating the formation of water molecules in dense molecular clouds. For this work, we have used a novel merged fast-beams apparatus which overlaps a beam of H3+ onto a beam of ground-term neutral O. Read More

To an adult, it's obvious that the day of someone's death is not precisely determined by the day of birth, but it's a very different story for a child. When the third named author was four years old he asked his father, the fifth named author: If two people are born on the same day, do they die on the same day? While this could easily be demonstrated through murder, such a proof would greatly diminish the possibility of teaching additional lessons, and thus a different approach was taken. With the help of the fourth named author they invented what we'll call \emph{the M\&M Game}: Given $k$ people, each simultaneously flips a fair coin, with each eating an M\&M on a head and not eating on a tail. Read More

Neighbor discovery is a fundamental task for wireless networks deployment. It is essential for setup and maintenance of networks and is typically a precondition for further communication. In this work we focus on passive discovery of networks operating in multi-channel environments, performed by listening for periodically transmitted beaconing messages. Read More

We consider a weighted random walk on the backbone of an oriented percolation cluster. We determine necessary conditions on the weights for Brownian scaling limits under the annealed and the quenched law. This model is a random walk in dynamic random environment (RWDRE), where the environment is mixing, non-Markovian and not elliptic. Read More

Recently, HTTP-Based Adaptive Streaming has become the de facto standard for video streaming over the Internet. It allows the client to adapt media characteristics to varying network conditions in order to maximize Quality of Experience (QoE). In the case of live streaming this task becomes particularly challenging. Read More

Recently, the way people consume video content has been undergoing a dramatic change. Plain TV sets, that have been the center of home entertainment for a long time, are losing grounds to Hybrid TV's, PC's, game consoles, and, more recently, mobile devices such as tablets and smartphones. The new predominant paradigm is: watch what I want, when I want, and where I want. Read More

Systems that evolve over time and follow mathematical laws as they do so, are called dynamical systems. Lymphocyte recovery and clinical outcomes in 41 allograft recipients conditioned using anti-thymocyte globulin (ATG) and 4.5 Gray total-body-irradiation were studied to determine if immune reconstitution could be described as a dynamical system. Read More

We have investigated the chemistry of ${\rm C + H_3^+}$ forming CH$^+$, CH$_2^+$, and CH$_3^+$. These reactions are believed to be some of the key gas-phase astrochemical processes initiating the formation of organic molecules in molecular clouds. For this work we have constructed a novel merged fast-beams apparatus which overlaps a beam of molecular ions onto a beam of ground-term neutral atoms. Read More

We provide a general formula for the eigenvalue density of large random $N\times N$ matrices of the form $A = M + LJR$, where $M$, $L$ and $R$ are arbitrary deterministic matrices and $J$ is a random matrix of zero-mean independent and identically distributed elements. For $A$ nonnormal, the eigenvalues do not suffice to specify the dynamics induced by $A$, so we also provide general formulae for the transient evolution of the magnitude of activity and frequency power spectrum in an $N$-dimensional linear dynamical system with a coupling matrix given by $A$. These quantities can also be thought of as characterizing the stability and the magnitude of the linear response of a nonlinear network to small perturbations about a fixed point. Read More

The electronic properties of Cu$_{0.07}$Bi$_{2}$Se$_{3}$ have been investigated using Shubnikov-de Haas and optical reflectance measurements. Quantum oscillations reveal a bulk, three-dimensional Fermi surface with anisotropy $k^{c}_{F}/k^{ab}_{F}\approx$ 2 and a modest increase in free-carrier concentration and in scattering rate with respect to the undoped Bi$_{2}$Se$_{3}$, also confirmed by reflectivity data. Read More

Reflection and transmission as a function of temperature (5--300 K) have been measured on single crystals of the multiferroic compound FeTe$_{2}$O$_{5}$Br utilizing light spanning the far infrared to the visible portions of the electromagnetic spectrum. The complex dielectric function and optical properties were obtained via Kramers-Kronig analysis and by fits to a Drude-Lortentz model. Analysis of the anisotropic excitation spectra via Drude-Lorentz fitting and lattice dynamical calculations have lead to the observation of all 52 IR-active modes predicted in the $ac$ plane and 43 or the 53 modes predicted along the b axis of the monoclinic cell. Read More

We performed temperature dependent infrared spectroscopy measurements on BiTeI single crystals, which exhibit large Rashba spin-splitting. Similar to a previous optical study, we found electronic excitations in good agreement with spin-split electronic bands. In addition, we report a low energy intraband transition with an onset energy of about 40 meV and an unexpectedly large number of vibrational modes in the far-infrared spectral region. Read More

Nonparametric Bayesian models are often based on the assumption that the objects being modeled are exchangeable. While appropriate in some applications (e.g. Read More

Infrared reflection and transmission as a function of temperature have been measured on single crystals of Cu$_{3}$Bi(SeO$_{3}$)$_{2}$O$_{2}$Cl. The complex dielectric function and optical properties along all three principal axes of the orthorhombic cell were obtained via Kramers-Kronig analysis and by fits to a Drude-Lorentz model. Below 115 K, 16 additional modes (8(E$\parallel\hat{a}$)+6(E$\parallel\hat{b}$)+2(E$\parallel\hat{c}$)) appear in the phonon spectra; however, powder x-ray diffraction measurements do not detect a new structure at 85 K. Read More

We study a rate-model neural network composed of excitatory and inhibitory neurons in which neuronal input-output functions are power laws with a power greater than 1, as observed in primary visual cortex. This supralinear input-output function leads to supralinear summation of network responses to multiple inputs for weak inputs. We show that for stronger inputs, which would drive the excitatory subnetwork to instability, the network will dynamically stabilize provided feedback inhibition is sufficiently strong. Read More

This paper presents a multigrid algorithm for the computation of the rank-R canonical decomposition of a tensor for low rank R. Standard alternating least squares (ALS) is used as the relaxation method. Transfer operators and coarse-level tensors are constructed in an adaptive setup phase based on multiplicative correction and on Bootstrap algebraic multigrid. Read More

Using a merged beams apparatus we have measured the associative detachment (AD) reaction of ${\rm H}^- + {\rm H} \to {\rm H}_2 + e^-$ for relative collision energies up to \textit{E}$_{\rm{r}} \leq 4.83$ eV. These data extend above the 1 eV limit of our earlier results. Read More

Loopy belief propagation performs approximate inference on graphical models with loops. One might hope to compensate for the approximation by adjusting model parameters. Learning algorithms for this purpose have been explored previously, and the claim has been made that every set of locally consistent marginals can arise from belief propagation run on a graphical model. Read More

Reflection and transmission as a function of temperature have been measured on a single crystal of the magnetoelectric ferrimagnetic compound Cu$_{2}$OSeO$_{3}$ utilizing light spanning the far infrared to the visible portions of the electromagnetic spectrum. The complex dielectric function and optical properties were obtained via Kramers-Kronig analysis and by fits to a Drude-Lortentz model. The fits of the infrared phonons show a magnetodielectric effect near the transition temperature ($T_{c}\sim 60$~K). Read More

Nano granular metallic iron (Fe) and titanium dioxide (TiO$_{2-\delta}$) were co-deposited on (100) lanthanum aluminate (LaAlO$_3$) substrates in a low oxygen chamber pressure using a pulsed laser ablation deposition (PLD) technique. The co-deposition of Fe and TiO$_2$ resulted in $\approx$ 10 nm metallic Fe spherical grains suspended within a TiO$_{2-\delta}$ matrix. The films show ferromagnetic behavior with a saturation magnetization of 3100 Gauss at room temperature. Read More

We report the results of our search for power-law electrical signals in the human brain, using subdural electrocorticographic recordings from the surface of the cortex. The power spectral density (PSD) of these signals has the power-law form $ P(f)\sim f^{-\chi} $ from 80 to 500 Hz. This scaling index $\chi = 4. Read More

Sensory neuroscience seeks to understand how the brain encodes natural environments. However, neural coding has largely been studied using simplified stimuli. In order to assess whether the brain's coding strategy depend on the stimulus ensemble, we apply a new information-theoretic method that allows unbiased calculation of neural filters (receptive fields) from responses to natural scenes or other complex signals with strong multipoint correlations. Read More

We analyse the non-linear, three-dimensional response of a gaseous, viscous protoplanetary disc to the presence of a planet of mass ranging from one Earth mass (1 M$_e$) to one Jupiter mass (1 M$_J$) by using the ZEUS hydrodynamics code. We determine the gas flow pattern, and the accretion and migration rates of the planet. The planet is assumed to be in a fixed circular orbit about the central star. Read More

We use three-dimensional magnetohydrodynamical (MHD) simulations to study the formation of a corona above an initially weakly magnetized, isothermal accretion disk. We also describe a modification to time-explicit numerical algorithms for MHD which enables us to evolve highly stratified disks for many orbital times. We find that MHD turbulence driven by the magnetorotational instability (MRI) produces strong amplification of weak fields within two scale heights of the disk midplane in a few orbital times. Read More

Results of a search for small-scale anisotropy in the cosmic microwave background (CMB) are presented. Observations were made at the South Pole using the Viper telescope, with a .26 degree (FWHM) beam and a passband centered at 40 GHz. Read More

Observations of the microwave sky using the Python telescope in its fifth season of operation at the Amundsen-Scott South Pole Station in Antarctica are presented. The system consists of a 0.75 m off-axis telescope instrumented with a HEMT amplifier-based radiometer having continuum sensitivity from 37-45 GHz in two frequency bands. Read More

Gryphon is a distributed computing paradigm for message brokering, which is the transferring of information in the form of streams of events from information providers to information consumers. This extended abstract outlines the major problems in message brokering and Gryphon's approach to solving them. Read More