-
Evolving symbolic density functionals
Authors:
He Ma,
Arunachalam Narayanaswamy,
Patrick Riley,
Li Li
Abstract:
Systematic development of accurate density functionals has been a decades-long challenge for scientists. Despite the emerging application of machine learning (ML) in approximating functionals, the resulting ML functionals usually contain more than tens of thousands parameters, which makes a huge gap in the formulation with the conventional human-designed symbolic functionals. We propose a new fram…
▽ More
Systematic development of accurate density functionals has been a decades-long challenge for scientists. Despite the emerging application of machine learning (ML) in approximating functionals, the resulting ML functionals usually contain more than tens of thousands parameters, which makes a huge gap in the formulation with the conventional human-designed symbolic functionals. We propose a new framework, Symbolic Functional Evolutionary Search (SyFES), that automatically constructs accurate functionals in the symbolic form, which is more explainable to humans, cheaper to evaluate, and easier to integrate to existing density functional theory codes than other ML functionals. We first show that without prior knowledge, SyFES reconstructed a known functional from scratch. We then demonstrate that evolving from an existing functional $ω$B97M-V, SyFES found a new functional, GAS22 (Google Accelerated Science 22), that performs better for the majority of molecular types in the test set of Main Group Chemistry Database (MGCDB84). Our framework opens a new direction in leveraging computing power for the systematic development of symbolic density functionals.
△ Less
Submitted 23 August, 2022; v1 submitted 3 March, 2022;
originally announced March 2022.
-
Scientific Discovery by Generating Counterfactuals using Image Translation
Authors:
Arunachalam Narayanaswamy,
Subhashini Venugopalan,
Dale R. Webster,
Lily Peng,
Greg Corrado,
Paisan Ruamviboonsuk,
Pinal Bavishi,
Rory Sayres,
Abigail Huang,
Siva Balasubramanian,
Michael Brenner,
Philip Nelson,
Avinash V. Varadarajan
Abstract:
Model explanation techniques play a critical role in understanding the source of a model's performance and making its decisions transparent. Here we investigate if explanation techniques can also be used as a mechanism for scientific discovery. We make three contributions: first, we propose a framework to convert predictions from explanation techniques to a mechanism of discovery. Second, we show…
▽ More
Model explanation techniques play a critical role in understanding the source of a model's performance and making its decisions transparent. Here we investigate if explanation techniques can also be used as a mechanism for scientific discovery. We make three contributions: first, we propose a framework to convert predictions from explanation techniques to a mechanism of discovery. Second, we show how generative models in combination with black-box predictors can be used to generate hypotheses (without human priors) that can be critically examined. Third, with these techniques we study classification models for retinal images predicting Diabetic Macular Edema (DME), where recent work showed that a CNN trained on these images is likely learning novel features in the image. We demonstrate that the proposed framework is able to explain the underlying scientific mechanism, thus bridging the gap between the model's performance and human understanding.
△ Less
Submitted 19 July, 2020; v1 submitted 10 July, 2020;
originally announced July 2020.
-
It's easy to fool yourself: Case studies on identifying bias and confounding in bio-medical datasets
Authors:
Subhashini Venugopalan,
Arunachalam Narayanaswamy,
Samuel Yang,
Anton Geraschenko,
Scott Lipnick,
Nina Makhortova,
James Hawrot,
Christine Marques,
Joao Pereira,
Michael Brenner,
Lee Rubin,
Brian Wainger,
Marc Berndl
Abstract:
Confounding variables are a well known source of nuisance in biomedical studies. They present an even greater challenge when we combine them with black-box machine learning techniques that operate on raw data. This work presents two case studies. In one, we discovered biases arising from systematic errors in the data generation process. In the other, we found a spurious source of signal unrelated…
▽ More
Confounding variables are a well known source of nuisance in biomedical studies. They present an even greater challenge when we combine them with black-box machine learning techniques that operate on raw data. This work presents two case studies. In one, we discovered biases arising from systematic errors in the data generation process. In the other, we found a spurious source of signal unrelated to the prediction task at hand. In both cases, our prediction models performed well but under careful examination hidden confounders and biases were revealed. These are cautionary tales on the limits of using machine learning techniques on raw data from scientific experiments.
△ Less
Submitted 6 April, 2020; v1 submitted 12 December, 2019;
originally announced December 2019.
-
Image Detection and Digit Recognition to solve Sudoku as a Constraint Satisfaction Problem
Authors:
Aditya Narayanaswamy,
Yichuan Philip Ma,
Piyush Shrivastava
Abstract:
Sudoku is a puzzle well-known to the scientific community with simple rules of completion, which may require a com-plex line of reasoning. This paper addresses the problem of partitioning the Sudoku image into a 1-D array, recognizing digits from the array and representing it as a Constraint Sat-isfaction Problem (CSP). In this paper, we introduce new fea-ture extraction techniques for recognizing…
▽ More
Sudoku is a puzzle well-known to the scientific community with simple rules of completion, which may require a com-plex line of reasoning. This paper addresses the problem of partitioning the Sudoku image into a 1-D array, recognizing digits from the array and representing it as a Constraint Sat-isfaction Problem (CSP). In this paper, we introduce new fea-ture extraction techniques for recognizing digits, which are used with our benchmark classifiers in conjunction with the CSP algorithms to provide performance assessment. Experi-mental results show that application of CSP techniques can decrease the solution's search time by eliminating incon-sistent values from the search space.
△ Less
Submitted 25 May, 2019;
originally announced May 2019.
-
Thermal radiative energy exchange between a closely-spaced linear chain of spheres and its environment
Authors:
Braden Czapla,
Arvind Narayanaswamy
Abstract:
In this work, we present expressions for radiative heat transfer between pairs of spheres in a linear chain and between individual spheres and their environment. The expressions are valid for coated spheres of arbitrary size, spacing, and isotropic optical properties. The spheres may be small and closely-spaced, which violates the assumptions foundational to classical radiative transfer. We valida…
▽ More
In this work, we present expressions for radiative heat transfer between pairs of spheres in a linear chain and between individual spheres and their environment. The expressions are valid for coated spheres of arbitrary size, spacing, and isotropic optical properties. The spheres may be small and closely-spaced, which violates the assumptions foundational to classical radiative transfer. We validate our results against existing formulations of radiative heat transfer, namely the thermal discrete dipole and boundary element methods. Our results have important implications for the modeling and interpretation of near-field radiative heat transfer experiments between spherical bodies.
△ Less
Submitted 27 December, 2018;
originally announced December 2018.
-
Predicting optical coherence tomography-derived diabetic macular edema grades from fundus photographs using deep learning
Authors:
Avinash Varadarajan,
Pinal Bavishi,
Paisan Raumviboonsuk,
Peranut Chotcomwongse,
Subhashini Venugopalan,
Arunachalam Narayanaswamy,
Jorge Cuadros,
Kuniyoshi Kanai,
George Bresnick,
Mongkol Tadarati,
Sukhum Silpa-archa,
Jirawut Limwattanayingyong,
Variya Nganthavee,
Joe Ledsam,
Pearse A Keane,
Greg S Corrado,
Lily Peng,
Dale R Webster
Abstract:
Diabetic eye disease is one of the fastest growing causes of preventable blindness. With the advent of anti-VEGF (vascular endothelial growth factor) therapies, it has become increasingly important to detect center-involved diabetic macular edema (ci-DME). However, center-involved diabetic macular edema is diagnosed using optical coherence tomography (OCT), which is not generally available at scre…
▽ More
Diabetic eye disease is one of the fastest growing causes of preventable blindness. With the advent of anti-VEGF (vascular endothelial growth factor) therapies, it has become increasingly important to detect center-involved diabetic macular edema (ci-DME). However, center-involved diabetic macular edema is diagnosed using optical coherence tomography (OCT), which is not generally available at screening sites because of cost and workflow constraints. Instead, screening programs rely on the detection of hard exudates in color fundus photographs as a proxy for DME, often resulting in high false positive or false negative calls. To improve the accuracy of DME screening, we trained a deep learning model to use color fundus photographs to predict ci-DME. Our model had an ROC-AUC of 0.89 (95% CI: 0.87-0.91), which corresponds to a sensitivity of 85% at a specificity of 80%. In comparison, three retinal specialists had similar sensitivities (82-85%), but only half the specificity (45-50%, p<0.001 for each comparison with model). The positive predictive value (PPV) of the model was 61% (95% CI: 56-66%), approximately double the 36-38% by the retinal specialists. In addition to predicting ci-DME, our model was able to detect the presence of intraretinal fluid with an AUC of 0.81 (95% CI: 0.81-0.86) and subretinal fluid with an AUC of 0.88 (95% CI: 0.85-0.91). The ability of deep learning algorithms to make clinically relevant predictions that generally require sophisticated 3D-imaging equipment from simple 2D images has broad relevance to many other applications in medical imaging.
△ Less
Submitted 31 July, 2019; v1 submitted 18 October, 2018;
originally announced October 2018.
-
Near-field thermal radiative transfer between two coated spheres
Authors:
Braden Czapla,
Arvind Narayanaswamy
Abstract:
In this work, we present an expression for the near-field thermal radiative transfer between two spheres with an arbitrary numbers of coatings. We numerically demonstrate that the spectrum of heat transfer between layered spheres exhibits novel features due to the newly introduced interfaces between coatings and cores. These features include broad super-Planckian peaks at non-resonant frequencies…
▽ More
In this work, we present an expression for the near-field thermal radiative transfer between two spheres with an arbitrary numbers of coatings. We numerically demonstrate that the spectrum of heat transfer between layered spheres exhibits novel features due to the newly introduced interfaces between coatings and cores. These features include broad super-Planckian peaks at non-resonant frequencies and near-field selective emission between metallic spheres with polar material coatings. Spheres with cores and coatings of two different polar materials are also shown to exceed the total conductance of homogeneous spheres in some cases.
△ Less
Submitted 7 September, 2017; v1 submitted 3 March, 2017;
originally announced March 2017.
-
A Green's function formalism of energy and momentum transfer in fluctuational electrodynamics
Authors:
Arvind Narayanaswamy,
Yi Zheng
Abstract:
Radiative energy and momentum transfer due to fluctuations of electromagnetic fields arising due to temperature difference between objects is described in terms of the cross-spectral densities of the electromagnetic fields. We derive relations between thermal non-equilibrium contributions to energy and momentum transfer and surface integrals of tangential components of the dyadic Green's functions…
▽ More
Radiative energy and momentum transfer due to fluctuations of electromagnetic fields arising due to temperature difference between objects is described in terms of the cross-spectral densities of the electromagnetic fields. We derive relations between thermal non-equilibrium contributions to energy and momentum transfer and surface integrals of tangential components of the dyadic Green's functions of the vector Helmholtz equation. The expressions derived here are applicable to objects of arbitrary shapes, dielectric functions, as well as magnetic permeabilities. For the case of radiative transfer, we derive expressions for the generalized transmissivity and generalized conductance that are shown to obey reciprocity and agree with theory of black body radiative transfer in the appropriate limit.
△ Less
Submitted 3 February, 2013;
originally announced February 2013.
-
Entropy density and entropy flux in near-field thermal radiation
Authors:
Arvind Narayanaswamy,
Yi Zheng
Abstract:
We propose a method to evaluate the entropy density and entropy flux in a vacuum gap between two half-spaces that takes into account influence of near-field effects, i.e., interference, diffraction, and tunneling of waves. The method developed is used to determine the maximum work that can be extracted through near-field radiative transfer between two half-spaces at different temperatures.
We propose a method to evaluate the entropy density and entropy flux in a vacuum gap between two half-spaces that takes into account influence of near-field effects, i.e., interference, diffraction, and tunneling of waves. The method developed is used to determine the maximum work that can be extracted through near-field radiative transfer between two half-spaces at different temperatures.
△ Less
Submitted 29 October, 2012;
originally announced October 2012.
-
Lifshitz theory of van der Waals pressure in dissipative media
Authors:
Yi Zheng,
Arvind Narayanaswamy
Abstract:
We derive a first--principles method of determining the van der Waals or Casimir pressure in a dissipative and dispersive planar multilayered system by calculating the Maxwell stress tensor in a fictitious layer of vacuum, that is eventually made to vanish, introduced in the structure. This is illustrated by calculating the van der Waals pressure in a thin film with dissipative properties embedded…
▽ More
We derive a first--principles method of determining the van der Waals or Casimir pressure in a dissipative and dispersive planar multilayered system by calculating the Maxwell stress tensor in a fictitious layer of vacuum, that is eventually made to vanish, introduced in the structure. This is illustrated by calculating the van der Waals pressure in a thin film with dissipative properties embedded between two semi--infinite media.
△ Less
Submitted 27 November, 2010; v1 submitted 24 November, 2010;
originally announced November 2010.
-
Proximity Effects in Radiative Transfer
Authors:
Karthik Sasihithlu,
Arvind Narayanaswamy
Abstract:
Though the dependence of near-field radiative transfer on the gap between two planar objects is well understood, that between curved objects is still unclear. We show, based on the analysis of the surface polariton mediated radiative transfer between two spheres of equal radii $R$ and minimum gap $d$, that the near--field radiative transfer scales as $R/d$ as $d/R \rightarrow 0$ and as $\ln(R/d)$…
▽ More
Though the dependence of near-field radiative transfer on the gap between two planar objects is well understood, that between curved objects is still unclear. We show, based on the analysis of the surface polariton mediated radiative transfer between two spheres of equal radii $R$ and minimum gap $d$, that the near--field radiative transfer scales as $R/d$ as $d/R \rightarrow 0$ and as $\ln(R/d)$ for larger values of $d/R$ up to the far--field limit. We propose a modified form of the proximity approximation to predict near--field radiative transfer between curved objects from simulations of radiative transfer between planar surfaces.
△ Less
Submitted 5 October, 2010;
originally announced October 2010.
-
Dyadic Green's functions and electromagnetic local density of states
Authors:
Arvind Narayanaswamy,
Gang Chen
Abstract:
A formal proof to relate the concept of electromagnetic local density of states (LDOS) to the electric and magnetic dyadic Green's functions is provided. The expression for LDOS is obtained by relating the electromagnetic energy density at any location in a medium at uniform temperature $T$ to the electric and magnetic dyadic Green's functions. With this the concept of LDOS is also extended to m…
▽ More
A formal proof to relate the concept of electromagnetic local density of states (LDOS) to the electric and magnetic dyadic Green's functions is provided. The expression for LDOS is obtained by relating the electromagnetic energy density at any location in a medium at uniform temperature $T$ to the electric and magnetic dyadic Green's functions. With this the concept of LDOS is also extended to material media. The LDOS is split into two terms -- one that originates from the energy density in an infinite, homogeneous medium and the other that takes into account scattering from inhomogenieties. The second part can always be defined unambiguously, even in lossy materials. For lossy materials, the first part is finite only if spatial dispersion is taken into account.
△ Less
Submitted 3 September, 2009;
originally announced September 2009.
-
Near-field radiative heat transfer between a sphere and a substrate
Authors:
Arvind Narayanaswamy,
Sheng Shen,
Gang Chen
Abstract:
Near-field force and energy exchange between two objects due to quantum electrodynamic fluctuations give rise to interesting phenomena such as Casimir and van der Waals forces, and thermal radiative transfer exceeding Planck's theory of blackbody radiation. Although significant progress has been made in the past on the precise measurement of Casimir force related to zero-point energy, experiment…
▽ More
Near-field force and energy exchange between two objects due to quantum electrodynamic fluctuations give rise to interesting phenomena such as Casimir and van der Waals forces, and thermal radiative transfer exceeding Planck's theory of blackbody radiation. Although significant progress has been made in the past on the precise measurement of Casimir force related to zero-point energy, experimental demonstration of near-field enhancement of radiative heat transfer is difficult. In this work, we present a sensitive technique of measuring near-field radiative transfer between a microsphere and a substrate using a bi-material atomic force microscope (AFM) cantilever, resulting in "heat transfer-distance" curves. Measurements of radiative transfer between a sphere and a flat substrate show the presence of strong near-field effects resulting in enhancement of heat transfer over the predictions of the Planck blackbody radiation theory.
△ Less
Submitted 3 September, 2009;
originally announced September 2009.
-
Thermal near-field radiative transfer between two spheres
Authors:
Arvind Narayanaswamy,
Gang Chen
Abstract:
Radiative energy transfer between closely spaced bodies is known to be significantly larger than that predicted by classical radiative transfer because of tunneling due to evanescent waves. Theoretical analysis of near--field radiative transfer is mainly restricted to radiative transfer between two half--spaces or spheres treated in the dipole approximation (very small sphere) or proximity force…
▽ More
Radiative energy transfer between closely spaced bodies is known to be significantly larger than that predicted by classical radiative transfer because of tunneling due to evanescent waves. Theoretical analysis of near--field radiative transfer is mainly restricted to radiative transfer between two half--spaces or spheres treated in the dipole approximation (very small sphere) or proximity force approximation (radius of sphere much greater than the gap). Sphere--sphere or sphere--plane configurations beyond the dipole approximation or proximity force approximation have not been attempted. In this work, the radiative energy transfer between two adjacent non--overlap** spheres of arbitrary diameters and gaps is analyzed numerically. For spheres of small diameter (compared to the wavelength), the results coincide with the dipole approximation. We see that the proximity force approximation is not valid for spheres with diameters much larger than the gap, even though this approximation is well established for calculating forces. From the numerical results, a regime map is constructed based on two non--dimensional length scales for the validity of different approximations.
△ Less
Submitted 3 September, 2009;
originally announced September 2009.
