arXiv subjects

David Liu

Publications and source records attributed to David Liu.

48 records · Page 3Linked to original sources

arXiv2020

Evolving Antennas for Ultra-High Energy Neutrino Detection

Evolutionary algorithms borrow from biology the concepts of mutation and selection in order to evolve optimized solutions to known problems. The GENETIS collaboration is developing genetic algorithms for designing antennas that are more sensitive to ultra-high energy neutrino induced radio pulses than current designs. There are three aspects of this investigation. The first is to evolve simple wire antennas to test the concept and different algorithms. Second, optimized antenna response patterns are evolved for a given array geometry. Finally, antennas themselves are evolved using neutrino sensitivity as a measure of fitness. This is achieved by integrating the XFdtd finite-difference time-domain modeling program with simulations of neutrino experiments.

astro-ph.IM↗

arXiv2017

Automatic Vertebra Labeling in Large-Scale 3D CT using Deep Image-to-Image Network with Message Passing and Sparsity Regularization

Automatic localization and labeling of vertebra in 3D medical images plays an important role in many clinical tasks, including pathological diagnosis, surgical planning and postoperative assessment. However, the unusual conditions of pathological cases, such as the abnormal spine curvature, bright visual imaging artifacts caused by metal implants, and the limited field of view, increase the difficulties of accurate localization. In this paper, we propose an automatic and fast algorithm to localize and label the vertebra centroids in 3D CT volumes. First, we deploy a deep image-to-image network (DI2IN) to initialize vertebra locations, employing the convolutional encoder-decoder architecture together with multi-level feature concatenation and deep supervision. Next, the centroid probability maps from DI2IN are iteratively evolved with the message passing schemes based on the mutual relation of vertebra centroids. Finally, the localization results are refined with sparsity regularization. The proposed method is evaluated on a public dataset of 302 spine CT volumes with various pathologies. Our method outperforms other state-of-the-art methods in terms of localization accuracy. The run time is around 3 seconds on average per case. To further boost the performance, we retrain the DI2IN on additional 1000+ 3D CT volumes from different patients. To the best of our knowledge, this is the first time more than 1000 3D CT volumes with expert annotation are adopted in experiments for the anatomic landmark detection tasks. Our experimental results show that training with such a large dataset significantly improves the performance and the overall identification rate, for the first time by our knowledge, reaches 90 %.

cs.CV↗

arXiv2017

Visual Search at Pinterest

We demonstrate that, with the availability of distributed computation platforms such as Amazon Web Services and open-source tools, it is possible for a small engineering team to build, launch and maintain a cost-effective, large-scale visual search system with widely available tools. We also demonstrate, through a comprehensive set of live experiments at Pinterest, that content recommendation powered by visual search improve user engagement. By sharing our implementation details and the experiences learned from launching a commercial visual search engines from scratch, we hope visual search are more widely incorporated into today's commercial applications.

cs.CV↗

arXiv2017

Symmetry, stability, and computation of degenerate lasing modes

We present a general method to obtain the stable lasing solutions for the steady-state ab-initio lasing theory (SALT) for the case of a degenerate symmetric laser in two dimensions (2d). We find that under most regimes (with one pathological exception), the stable solutions are clockwise and counterclockwise circulating modes, generalizing previously known results of ring lasers to all 2d rotational symmetry groups. Our method uses a combination of semi-analytical solutions close to lasing threshold and numerical solvers to track the lasing modes far above threshold. Near threshold, we find closed-form expressions for both circulating modes and other types of lasing solutions as well as for their linearized Maxwell--Bloch eigenvalues, providing a simple way to determine their stability without having to do a full nonlinear numerical calculation. Above threshold, we show that a key feature of the circulating mode is its "chiral" intensity pattern, which arises from spontaneous symmetry-breaking of mirror symmetry, and whose symmetry group requires that the degeneracy persists even when nonlinear effects become important. Finally, we introduce a numerical technique to solve the degenerate SALT equations far above threshold even when spatial discretization artificially breaks the degeneracy.

physics.optics↗

arXiv2016

Interaction-induced mode switching in steady-state microlasers

We demonstrate that due to strong modal interactions through cross-gain saturation, the onset of a new lasing mode can switch off an existing mode via a negative power slope. In this process of interaction-induced mode switching (IMS) the two involved modes maintain their identities, i.e. they do not change their spatial field patterns or lasing frequencies. For a fixed pump profile, a simple analytic criterion for the occurrence of IMS is given in terms of their self- and cross-interaction coefficients and non-interacting thresholds, which is verified for the example of a two-dimensional microdisk laser. When the spatial pump profile is varied as the pump power is increased, IMS can be induced even when it would not occur with a fixed pump profile, as we show for two coupled laser cavities. Our findings apply to steady-state lasing and are hence different from dynamical mode switching or hopping. IMS may have potential applications in robust and flexible all-optical switching.

physics.optics↗

arXiv2015

Human Curation and Convnets: Powering Item-to-Item Recommendations on Pinterest

This paper presents Pinterest Related Pins, an item-to-item recommendation system that combines collaborative filtering with content-based ranking. We demonstrate that signals derived from user curation, the activity of users organizing content, are highly effective when used in conjunction with content-based ranking. This paper also demonstrates the effectiveness of visual features, such as image or object representations learned from convnets, in improving the user engagement rate of our item-to-item recommendation system.

cs.CV↗

arXiv2015

Ab-initio multimode linewidth theory for arbitrary inhomogeneous laser cavities

We present a multimode laser-linewidth theory for arbitrary cavity structures and geometries that contains nearly all previously known effects and also finds new nonlinear and multimode corrections, e.g. a bad-cavity correction to the Henry $α$ factor and a multimode Schawlow--Townes relation (each linewidth is proportional to a sum of inverse powers of all lasing modes). Our theory produces a quantitatively accurate formula for the linewidth, with no free parameters, including the full spatial degrees of freedom of the system. Starting with the Maxwell--Bloch equations, we handle quantum and thermal noise by introducing random currents whose correlations are given by the fluctuation--dissipation theorem. We derive coupled-mode equations for the lasing-mode amplitudes and obtain a formula for the linewidths in terms of simple integrals over the steady-state lasing modes.

physics.optics↗

arXiv2013

Pebbling Arguments for Tree Evaluation

The Tree Evaluation Problem was introduced by Cook et al. in 2010 as a candidate for separating P from L and NL. The most general space lower bounds known for the Tree Evaluation Problem require a semantic restriction on the branching programs and use a connection to well-known pebble games to generate a bottleneck argument. These bounds are met by corresponding upper bounds generated by natural implementations of optimal pebbling algorithms. In this paper we extend these ideas to a variety of restricted families of both deterministic and non-deterministic branching programs, proving tight lower bounds under these restricted models. We also survey and unify known lower bounds in our "pebbling argument" framework.

cs.CC↗

arXiv2013

Transformation Inverse Design

We present a new technique for the design of transformation-optics devices based on large-scale optimization to achieve the optimal effective isotropic dielectric materials within prescribed index bounds, which is computationally cheap because transformation optics circumvents the need to solve Maxwell's equations at each step. We apply this technique to the design of multimode waveguide bends (realized experimentally in a previous paper) and mode squeezers, in which all modes are transported equally without scattering. In addition to the optimization, a key point is the identification of the correct boundary conditions to ensure reflectionless coupling to untransformed regions while allowing maximum flexibility in the optimization. Many previous authors in transformation optics used a certain kind of quasiconformal map which overconstrained the problem by requiring that the entire boundary shape be specified a priori while at the same time underconstraining the problem by employing "slipping" boundary conditions that permit unwanted interface reflections.

physics.optics↗

arXiv2012

Universal Witnesses for State Complexity of Boolean Operations and Concatenation Combined with Star

We study the state complexity of boolean operations and product (concatenation, catenation) combined with star. We derive tight upper bounds for the symmetric differences and differences of two languages, one or both of which are starred, and for the product of two starred languages. We prove that the previously discovered bounds for the union and the intersection of languages with one or two starred arguments, for the product of two languages one of which is starred, and for the star of the product of two languages can all be met by the recently introduced universal witnesses and their variants.

cs.FL↗

arXiv2012

Universal Witnesses for State Complexity of Basic Operations Combined with Reversal

We study the state complexity of boolean operations, concatenation and star with one or two of the argument languages reversed. We derive tight upper bounds for the symmetric differences and differences of such languages. We prove that the previously discovered bounds for union, intersection, concatenation and star of such languages can all be met by the recently introduced universal witnesses and their variants.

cs.FL↗

arXiv2012

Syntactic Complexity of Finite/Cofinite, Definite, and Reverse Definite Languages

We study the syntactic complexity of finite/cofinite, definite and reverse definite languages. The syntactic complexity of a class of languages is defined as the maximal size of syntactic semigroups of languages from the class, taken as a function of the state complexity n of the languages. We prove that (n-1)! is a tight upper bound for finite/cofinite languages and that it can be reached only if the alphabet size is greater than or equal to (n-1)!-(n-2)!. We prove that the bound is also (n-1)! for reverse definite languages, but the minimal alphabet size is (n-1)!-2(n-2)!. We show that \lfloor e\cdot (n-1)!\rfloor is a lower bound on the syntactic complexity of definite languages, and conjecture that this is also an upper bound, and that the alphabet size required to meet this bound is \floor{e \cdot (n-1)!} - \floor{e \cdot (n-2)!}. We prove the conjecture for n\le 4.

cs.FL↗