-
Local Causal Discovery for Estimating Causal Effects
Authors:
Shantanu Gupta,
David Childers,
Zachary C. Lipton
Abstract:
Even when the causal graph underlying our data is unknown, we can use observational data to narrow down the possible values that an average treatment effect (ATE) can take by (1) identifying the graph up to a Markov equivalence class; and (2) estimating that ATE for each graph in the class. While the PC algorithm can identify this class under strong faithfulness assumptions, it can be computationa…
▽ More
Even when the causal graph underlying our data is unknown, we can use observational data to narrow down the possible values that an average treatment effect (ATE) can take by (1) identifying the graph up to a Markov equivalence class; and (2) estimating that ATE for each graph in the class. While the PC algorithm can identify this class under strong faithfulness assumptions, it can be computationally prohibitive. Fortunately, only the local graph structure around the treatment is required to identify the set of possible ATE values, a fact exploited by local discovery algorithms to improve computational efficiency. In this paper, we introduce Local Discovery using Eager Collider Checks (LDECC), a new local causal discovery algorithm that leverages unshielded colliders to orient the treatment's parents differently from existing methods. We show that there exist graphs where LDECC exponentially outperforms existing local discovery algorithms and vice versa. Moreover, we show that LDECC and existing algorithms rely on different faithfulness assumptions, leveraging this insight to weaken the assumptions for identifying the set of possible ATE values.
△ Less
Submitted 10 April, 2024; v1 submitted 15 February, 2023;
originally announced February 2023.
-
Efficient Online Estimation of Causal Effects by Deciding What to Observe
Authors:
Shantanu Gupta,
Zachary C. Lipton,
David Childers
Abstract:
Researchers often face data fusion problems, where multiple data sources are available, each capturing a distinct subset of variables. While problem formulations typically take the data as given, in practice, data acquisition can be an ongoing process. In this paper, we aim to estimate any functional of a probabilistic model (e.g., a causal effect) as efficiently as possible, by deciding, at each…
▽ More
Researchers often face data fusion problems, where multiple data sources are available, each capturing a distinct subset of variables. While problem formulations typically take the data as given, in practice, data acquisition can be an ongoing process. In this paper, we aim to estimate any functional of a probabilistic model (e.g., a causal effect) as efficiently as possible, by deciding, at each time, which data source to query. We propose online moment selection (OMS), a framework in which structural assumptions are encoded as moment conditions. The optimal action at each step depends, in part, on the very moments that identify the functional of interest. Our algorithms balance exploration with choosing the best action as suggested by current estimates of the moments. We propose two selection strategies: (1) explore-then-commit (OMS-ETC) and (2) explore-then-greedy (OMS-ETG), proving that both achieve zero asymptotic regret as assessed by MSE. We instantiate our setup for average treatment effect estimation, where structural assumptions are given by a causal graph and data sources may include subsets of mediators, confounders, and instrumental variables.
△ Less
Submitted 30 October, 2021; v1 submitted 20 August, 2021;
originally announced August 2021.
-
Reciprocal plasmonic metasurfaces: Theory and applications
Authors:
Yanzeng Li,
Micheal McLamb,
Serang Park,
Darrell Childers,
Glenn D. Boreman,
Tino Hofmann
Abstract:
A new configuration for metasurface construction is presented to achieve multi-functional capabilities including perfect absorption, bio/chem sensing, and surface-mode lasing. The reciprocal plasmonic metasurfaces discussed here are composed of two plasmonic surfaces of reciprocal geometries separated by a dielectric spacer. Compared to conventional metasurfaces this simple geometry exhibits an en…
▽ More
A new configuration for metasurface construction is presented to achieve multi-functional capabilities including perfect absorption, bio/chem sensing, and surface-mode lasing. The reciprocal plasmonic metasurfaces discussed here are composed of two plasmonic surfaces of reciprocal geometries separated by a dielectric spacer. Compared to conventional metasurfaces this simple geometry exhibits an enhanced optical performance. The discussed reciprocal metasurface design further enables effective structural optimization and allows for a simple and scalable fabrication. The physical principle and potential applications of the reciprocal plasmonic metasurfaces are demonstrated using numerical and analytical approaches.
△ Less
Submitted 1 May, 2020;
originally announced May 2020.
-
Estimating Treatment Effects with Observed Confounders and Mediators
Authors:
Shantanu Gupta,
Zachary C. Lipton,
David Childers
Abstract:
Given a causal graph, the do-calculus can express treatment effects as functionals of the observational joint distribution that can be estimated empirically. Sometimes the do-calculus identifies multiple valid formulae, prompting us to compare the statistical properties of the corresponding estimators. For example, the backdoor formula applies when all confounders are observed and the frontdoor fo…
▽ More
Given a causal graph, the do-calculus can express treatment effects as functionals of the observational joint distribution that can be estimated empirically. Sometimes the do-calculus identifies multiple valid formulae, prompting us to compare the statistical properties of the corresponding estimators. For example, the backdoor formula applies when all confounders are observed and the frontdoor formula applies when an observed mediator transmits the causal effect. In this paper, we investigate the over-identified scenario where both confounders and mediators are observed, rendering both estimators valid. Addressing the linear Gaussian causal model, we demonstrate that either estimator can dominate the other by an unbounded constant factor. Next, we derive an optimal estimator, which leverages all observed variables, and bound its finite-sample variance. We show that it strictly outperforms the backdoor and frontdoor estimators and that this improvement can be unbounded. We also present a procedure for combining two datasets, one with observed confounders and another with observed mediators. Finally, we evaluate our methods on both simulated data and the IHDP and JTPA datasets.
△ Less
Submitted 14 June, 2021; v1 submitted 26 March, 2020;
originally announced March 2020.
-
Fabrication of optical components with nm- to mm-scale critical features using three-dimensional direct laser writing
Authors:
Y. Li,
S. Park,
M. McLamb,
M. Lata,
D. Childers,
T. Hofmann
Abstract:
A powerful fabrication strategy based on three-dimensional direct laser writing for the rapid prototy** of opto-mechanical components with critical features ranging from several hundred nm to a few mm is demonstrated here. As an example, a simple optical fiber connector with optical and mechanical guides as well as integrated micro-optical elements with nano-structured surfaces is designed and f…
▽ More
A powerful fabrication strategy based on three-dimensional direct laser writing for the rapid prototy** of opto-mechanical components with critical features ranging from several hundred nm to a few mm is demonstrated here. As an example, a simple optical fiber connector with optical and mechanical guides as well as integrated micro-optical elements with nano-structured surfaces is designed and fabricated. In contrast to established three-dimensional direct laser writing, two different polymers are combined in the fabrication process in order to achieve a drastic reduction in fabrication time by substantially reducing the "optical tool path". A good agreement between the as-fabricated connector and nominal dimensions has been obtained. The developed approach allows the rapid prototy** of optomechanical components with multi-scale critical features. It is, therefore, envisioned to substantially accelerate the development cycle by integrating functional mechanical and optical elements in a single component.
△ Less
Submitted 27 September, 2019;
originally announced January 2020.
-
Broadband near-infrared antireflection coatings fabricated by three dimensional direct laser writing
Authors:
Y. Li,
D. B. Fullager,
E. Angelbello,
D. Childers,
G. Boreman,
T. Hofmann
Abstract:
Three-dimensional direct laser writing via two photon polymerization is used to fabricate anti-reflective structured surfaces composed of sub-wavelength conicoid features optimized to operate over a wide bandwidth in the near-infrared range from 3700 cm$^{-1}$ to 6600 cm$^{-1}$ (2.7 to 1.52 $μ$m). Analytic Bruggemann effective medium calculations are used to predict nominal geometric parameters su…
▽ More
Three-dimensional direct laser writing via two photon polymerization is used to fabricate anti-reflective structured surfaces composed of sub-wavelength conicoid features optimized to operate over a wide bandwidth in the near-infrared range from 3700 cm$^{-1}$ to 6600 cm$^{-1}$ (2.7 to 1.52 $μ$m). Analytic Bruggemann effective medium calculations are used to predict nominal geometric parameters such as the fill factor of the constitutive conicoid features of the anti-reflective structured surfaces presented here. The performance of the anti-reflective structured surfaces was investigated experimentally using infrared transmission measurements. An enhancement of the transmittance by 1.35\% to 2.14\% over a broadband spectral range from 3700 cm$ ^{-1} $ to 6600 cm$^{-1}$ (2.7 to 1.52 $μ$m) was achieved. We further report on finite-element-based reflection and transmission data using three-dimensional model geometries for comparison. A good agreement between experimental results and the finite-element-based numerical analysis is observed once as-fabricated deviations from the nominal conicoid forms are included in the model. Three-dimensional direct laser writing is demonstrated here as an efficient method for the fabrication and optimization of anti-reflective structured surfaces designed for the infrared spectral range.
△ Less
Submitted 29 September, 2017;
originally announced October 2017.
-
Empirical Determination of Convection Parameters in White Dwarfs I : Whole Earth Telescope Observations of EC14012-1446
Authors:
J. L. Provencal,
M. H. Montgomery,
A. Kanaan,
S. E. Thompson,
J. Dalessio,
H. L. Shipman,
D. Childers,
J. C. Clemens,
R. Rosen,
P. Henrique,
A. Bischoff-Kim,
W. Strickland,
D. Chandler,
B. Walter,
T. K. Watson,
B. Castanheira,
S. Wang,
G. Handler,
M. Wood,
S. Vennes,
P. Nemeth,
S. O. Kepler,
M. Reed,
A. Nitta,
S. J. Kleinman
, et al. (33 additional authors not shown)
Abstract:
We report on analysis of 308.3 hrs of high speed photometry targeting the pulsating DA white dwarf EC14012-1446. The data were acquired with the Whole Earth Telescope (WET) during the 2008 international observing run XCOV26. The Fourier transform of the light curve contains 19 independent frequencies and numerous combination frequencies. The dominant peaks are 1633.907, 1887.404, and 2504.897 micr…
▽ More
We report on analysis of 308.3 hrs of high speed photometry targeting the pulsating DA white dwarf EC14012-1446. The data were acquired with the Whole Earth Telescope (WET) during the 2008 international observing run XCOV26. The Fourier transform of the light curve contains 19 independent frequencies and numerous combination frequencies. The dominant peaks are 1633.907, 1887.404, and 2504.897 microHz. Our analysis of the combination amplitudes reveals that the parent frequencies are consistent with modes of spherical degree l=1. The combination amplitudes also provide m identifications for the largest amplitude parent frequencies. Our seismology analysis, which includes 2004--2007 archival data, confirms these identifications, provides constraints on additional frequencies, and finds an average period spacing of 41 s. Building on this foundation, we present nonlinear fits to high signal-to-noise light curves from the SOAR 4.1m, McDonald 2.1m, and KPNO 2m telescopes. The fits indicate a time-averaged convective response timescale of 99.4 +/- 17 s, a temperature exponent 85 +/- 6.2 and an inclination angle of 32.9 +/- 3.2 degrees. We present our current empirical map of the convective response timescale across the DA instability strip.
△ Less
Submitted 11 April, 2012;
originally announced April 2012.
-
An Extended Fatou-Shishikura inequality and wandering branch continua for polynomials
Authors:
Alexander Blokh,
Doug Childers,
Genadi Levin,
Lex Oversteegen,
Dierk Schleicher
Abstract:
Let $P$ be a polynomial of degree $d$ with Julia set $J_P$. Let $\widetilde N$ be the number of non-repelling cycles of $P$. By the famous Fatou-Shishikura inequality $\widetilde N\le d-1$. The goal of the paper is to improve this bound. The new count includes \emph{wandering collections of non-precritical branch continua}, i.e., collections of continua or points $Q_i\subset J_P$ \emph{all} of who…
▽ More
Let $P$ be a polynomial of degree $d$ with Julia set $J_P$. Let $\widetilde N$ be the number of non-repelling cycles of $P$. By the famous Fatou-Shishikura inequality $\widetilde N\le d-1$. The goal of the paper is to improve this bound. The new count includes \emph{wandering collections of non-precritical branch continua}, i.e., collections of continua or points $Q_i\subset J_P$ \emph{all} of whose images are pairwise disjoint, contain no critical points, and contain the limit sets of $\mathrm{eval}(Q_i)\ge 3$ external rays. Also, we relate individual cycles, which are either non-repelling or repelling with no periodic rays landing, to individual critical points that are recurrent in a weak sense.
A weak version of the inequality reads \[ \widetilde N+N_{irr}+χ+\sum_i (\mathrm{eval}(Q_i)-2) \le d-1 \] where $N_{irr}$ counts repelling cycles with no periodic rays landing at points in the cycle, $\{Q_i\}$ form a wandering collection $\mathcal{B}_\mathbb{C}$ of non-precritical branch continua, $χ=1$ if $\mathcal{B}_\mathbb{C}$ is non-empty, and $χ=0$ otherwise.
△ Less
Submitted 23 November, 2015; v1 submitted 6 January, 2010;
originally announced January 2010.
-
2006 Whole Earth Telescope Observations of GD358: A New Look at the Prototype DBV
Authors:
J. L. Provencal,
M. H. Montgomery,
A. Kanaan,
H. L. Shipman,
D. Childers,
A. Baran,
S. O. Kepler,
M. Reed,
A. Zhou,
J. Eggen,
T. K. Watson,
D. E. Winget,
S. E. Thompson,
B. Riaz,
A. Nitta,
S. J. Kleinman,
R. Crowe,
J. Slivkoff,
P. Sherard,
N. Purves,
P. Binder,
R. Knight,
S. -L. Kim,
Wen-** Chen,
M. Yang
, et al. (31 additional authors not shown)
Abstract:
We report on the analysis of 436.1 hrs of nearly continuous high-speed photometry on the pulsating DB white dwarf GD358 acquired with the Whole Earth Telescope (WET) during the 2006 international observing run, designated XCOV25. The Fourier transform (FT) of the light curve contains power between 1000 to 4000 microHz, with the dominant peak at 1234 microHz. We find 27 independent frequencies di…
▽ More
We report on the analysis of 436.1 hrs of nearly continuous high-speed photometry on the pulsating DB white dwarf GD358 acquired with the Whole Earth Telescope (WET) during the 2006 international observing run, designated XCOV25. The Fourier transform (FT) of the light curve contains power between 1000 to 4000 microHz, with the dominant peak at 1234 microHz. We find 27 independent frequencies distributed in 10 modes, as well as numerous combination frequencies. Our discussion focuses on a new asteroseismological analysis of GD358, incorporating the 2006 data set and drawing on 24 years of archival observations. Our results reveal that, while the general frequency locations of the identified modes are consistent throughout the years, the multiplet structure is complex and cannot be interpreted simply as l=1 modes in the limit of slow rotation. The high k multiplets exhibit significant variability in structure, amplitude and frequency. Any identification of the m components for the high k multiplets is highly suspect. The k=9 and 8 modes typically do show triplet structure more consistent with theoretical expectations. The frequencies and amplitudes exhibit some variability, but much less than the high k modes. Analysis of the k=9 and 8 multiplet splittings from 1990 to 2008 reveal a long-term change in multiplet splittings coinciding with the 1996 "sforzando" event, where GD358 dramatically altered its pulsation characteristics on a timescale of hours. We explore potential implications, including the possible connections between convection and/or magnetic fields and pulsations. We suggest future investigations, including theoretical investigations of the relationship between magnetic fields, pulsation, growth rates, and convection.
△ Less
Submitted 5 November, 2008;
originally announced November 2008.
-
Non-degenerate quadratic laminations
Authors:
A. Blokh,
D. Childers,
J. Mayer,
L. Oversteegen
Abstract:
We give a combinatorial criterion for a critical diameter to be compatible with a non-degenerate quadratic lamination.
We give a combinatorial criterion for a critical diameter to be compatible with a non-degenerate quadratic lamination.
△ Less
Submitted 11 September, 2008;
originally announced September 2008.
-
Wandering polygons and recurrent critical leaves
Authors:
Douglas K. Childers
Abstract:
Let T be a finite subset of the complex unit circle S^1, and define f: S^1 -> S^1 by f(z) = z^d. Let CH(T) denote the convex hull of T. If card(T) = N > 2, then CH(T) defines a polygon with N sides. The N-gon CH(T) is called a \emph{wandering N-gon} if for every two non-negative integers i \neq j, CH(f^i(T)) and CH(f^j(T)) are disjoint N-gons.
A non-degenerate chord of S^1 is said to be \emph{…
▽ More
Let T be a finite subset of the complex unit circle S^1, and define f: S^1 -> S^1 by f(z) = z^d. Let CH(T) denote the convex hull of T. If card(T) = N > 2, then CH(T) defines a polygon with N sides. The N-gon CH(T) is called a \emph{wandering N-gon} if for every two non-negative integers i \neq j, CH(f^i(T)) and CH(f^j(T)) are disjoint N-gons.
A non-degenerate chord of S^1 is said to be \emph{critical} if its two endpoints have the same image under f. Then for a critical chord, it is natural to define its (forward) orbit by the forward iterates of the endpoints. Similarly, call a critical chord \emph{recurrent} if one of its endpoints is recurrent under f. The main result of our study is that a wandering N-gon has at least N-1 recurrent critical chords in its limit set (defined in a natural way) having pairwise disjoint, infinite orbits.
Using this result, we are able to strengthen results of Blokh, Kiwi and Levin about wandering polygons of laminations. We also discuss some applications to the dynamics of polynomials. In particular, our study implies that if v is a wandering non-precritical vertex of a locally connected polynomial Julia set, then there exists at least ord(v)-1 recurrent critical points with pairwise disjoint orbits, all having the same omega-limit set as v. Thus, we likewise strengthen results about wandering vertices of polynomial Julia sets.
△ Less
Submitted 29 May, 2005;
originally announced May 2005.
