Abstract

This study offers an analytical estimation model for radiative scattering at nanoscale. The study focuses on isolated nanowires of arbitrary shape cross sections and uses predictive geometric features and statistical regression to model the wavelength-dependent light-particle interaction. This work proposes to estimate the radiative properties of nanowires based on engineered geometric features, potentially leading to new understandings of how the geometric attributes impact light scattering at nanoscale. A predictive model is designed and tested for estimating radiative scattering around nanowires. Random polygon-shaped cross sections with high degrees-of-freedom are chosen to train and test the models. The derived model can successfully explain scattering across out-sample synthetic plasmonic objects with a 90% R-squared metric.

1 Introduction

Optical characterization and design of subwavelength objects are challenging problems due to the complexity and computational cost of modeling. Nevertheless, the significant recent engineering advantages offered by the subwavelength particles in radiation applications such as cellular imaging [1], cancer therapy [2], optical antennas [3], photovoltaic cells [4], and more, have urged the optics and nanoscience researchers to come up with systematic ways for studying nanoscale radiation. Despite the growth in computational resources, optical nanoscale design remains an intense repetitive numerical process involving solving Maxwell's partial differential equations. Analytical solutions are widely missing except in limited cases involving well-shaped (spherical, cylindrical, etc.) isolated objects.

1.1 Summary of Contributions.

This study proposes an analytical model for the characterization and prediction of radiative response (scattering) around isolated nanowires. This is done in three steps: (1) training data generation: a large dataset of nanowires with arbitrary shape cross sections is generated as the training data. A random 2D polygon generation model is proposed and used to model arbitrary cross sections; (2) predictive feature extraction: we use a common set of geometric features (shape descriptors) to characterize the physical cross section shapes. These geometric features can be used as independent variables in a fitting framework to derive predictive models for the radiative response of nanoparticles in the face of incident light/radiation; and (3) model fitting: feature selection and regression techniques are used to obtain reliable predictive models. These models can reliably replace full-fidelity time-consuming electromagnetic simulations, and efficiently predict radiative scattering provided that the simulation parameters are close to those used in training. Otherwise, such models can still be useful in generalized applications with different settings if accuracy is not crucial, e.g., educational purposes or rapid (surrogate-model) optimizations. In addition, the provided analytical models can be helpful in more rigorous investigations of the effects of geometric shapes in scattering properties, e.g., in inverse design.

1.2 Limitations of the Study.

Our study is limited to 2D radiation and is therefore focused on nanowires (not nanoparticles). For data generation and shape description, we use randomly generated polygons. Every real-world 2D shape can be approximated to within a required error by a polygon of appropriate order.2 Intuitive features derived from polygon shapes can be as simple as area and eccentricity, and as complicated as high-order metrics of convexity, elongation, etc. The geometric shape features are computed using image characterization methods adopted from image recognition with nuances to incorporate the interaction of light and objects. The work is limited to metallic (silver) objects. A more comprehensive study shall include a larger set of material bases, both other plasmonic and dielectric materials, and generalizing the learning algorithms to include material properties as features (or normalization parameters). The work is also limited to isolated, compact, finite, homogenous particles.

1.3 Background and Motivation.

The optical response of isolated nanoparticles and nanowires to radiation is a highly complex function of the physical properties such as particle size, shape and material, as well as the characteristics of the incident light (e.g., wavelength, polarization, angle, etc.). The radiative response profile (i.e., spectral reflectance, absorbance, scattering) can be characterized by multiple metrics, including the response strength, bandgap, stop band, bandwidth, resonance/peak frequency, number, location and range of peaks, etc. These characteristics are intertwined and adjustable via the choice of geometry and material. The tunability of the radiative spectrum makes the geometry design an attractive method for unique devices that utilize nanoparticles and nanowires [57]. In some simplified cases, e.g., when the particle size is significantly smaller than the wavelength, the quasi-static assumption holds and the impact of the particle shape can be approximated via various shape factors [3,810]. Mie’s theory provides an analytical solution to Maxwell's equations in the case of spherical shapes [11], with a few limited extensions to non-spherical shapes [12,13].

Studying non-spherical/non-circular objects with sizes comparable to the wavelength requires using computational Maxwell's equation solvers, such as the finite element method, finite difference method, and discrete dipole approximation. Recently, shape-dependent optical response of nanoparticles has been studied by many researchers by considering common shapes such as triangles, cylinders, spheroids, and cuboids [1419], as well as more exotic forms such as bipyramids [20], star and flower shapes [21], bowl and dumbbell shapes [22]. The research on shape-dependent optical properties has led to novel approaches for shape exploration in radiation. For instance, several studies have utilized structural topology optimization to determine suitable nanoscale shapes for optical problems [23,24]. In Refs. [2530], the efficiency of solar cells was optimized using light-trapping structures with the help of topology optimization. In Ref. [31], photonic crystal waveguides were optimized for desired dispersion properties using similar approaches.

Data-driven approaches are another class of techniques that have recently gained popularity in nano-photonics study and design, with the aim of alleviating the computational burden of modeling [3243]. The shape geometry in these studies is expressed either by some mathematical parametrization, such as side lengths [44,45] and vertex coordinates in the case of polygons [46] or by pixel representation, analogous to the field of image recognition [47,48]. Though limited, the effect of the shape features on the optical properties has also been studied in certain cases (e.g., Refs. [49,50]). For example, in Ref. [49], the emissivity of particles with four different shapes (rectangular and triangular prisms, sphere, cylinder) was predicted using simple geometric features, namely area-to-volume ratio, shortest, middle, and longest dimensions. Nevertheless, the generality of particle shapes and optical property prediction based on intuitive geometric features are widely missing in the existing literature. We can therefore assert that the state-of-the-art data-driven modeling of nanoscale radiation and scattering is limited and incomplete. This work is a preliminary attempt to fill some of that knowledge gap.

The rest of the paper is organized as follows: First, a quick overview of the light scattering theory is presented. Then, the study framework for nanowires with arbitrary shape cross sections is elaborated. This includes the shape generation methodology, predictive geometric features for shape characterization, feature selection, and regression analysis. Numerical results are presented and discussed using test cases at the end and concluded.

2 Light Scattering at Nanoscale

Particles that are exposed to light absorb or scatter the incoming energy. Light scattering results from the interaction of the electromagnetic field with objects of different permittivity distributions and interfaces between these objects [51]. Scattering and absorption together comprise extinction. A common metric to measure light extinction is the optical cross section, defined as the power extincted, scattered, or absorbed by the object normalized by the irradiance, i.e.,
Pe=Ps+Pa
(1)
C=PI0,{a,e,s}
(2)
where P is power, I0 is irradiance, and C is the optical cross section, respectively. Subscripts e, s, a stand for extinction, scattering, and absorption, respectively. These cross section quantities are wavelength-dependent values that can differ from the geometric cross sections, and up to orders of magnitude larger with selective designs, particularly at small dimensions.
Light material interaction is explained by the electromagnetic theory and Maxwell's equations, which are partial derivative equations of electric, magnetic, and displacement fields. Under certain circumstances, the solution can be greatly simplified, rendering explicit solutions. For example, considering the size of an object is significantly smaller than the wavelength of light, electrostatic approximations hold and the scattering is primarily in the form of Rayleigh scattering. Another condition that results in an explicit solution is the symmetry in the object's cross section. Mie's theory provides analytical solutions for sphere and cylinder like objects [52,53]. When the particle size is comparable to or smaller than the wavelength, and the shape is non-spherical, computational methods are required for accurate electromagnetic modeling. Finite difference time domain (FDTD) is one of the widely used methods to model electromagnetic field components due to its simplicity and accuracy [5458]. FDTD solves Maxwell's equations on a discrete spatial and temporal grid, composed of the collection of so-called Yee's cells.3 Yee's cell involves electric and magnetic field vectors along edges and perpendicular to the faces, respectively. Once the electric and magnetic fields are calculated on the solution domain, the radiation power crossing a plane S is computable using
P=12SRe(E×H)dS
(3)
where E and H are electric and magnetic fields [59]. In calculating the electromagnetic field, the incident light is considered unpolarized, i.e., the average of the transverse magnetic (TM) and transverse electric (TE) modes is taken. The optical cross sections are then calculated using Eq. (2).

3 Model Framework Description

This study aims to establish a relationship between the shape of nanowires cross sections and its radiative scattering using a statistical approach with engineered geometric features as inputs. The procedure is summarized in Fig. 1. First, random polygons are generated using a methodology described in the next section. Then electromagnetic simulations are performed to find target outputs (scattering cross section) for the generated shapes. Meanwhile, the values of geometric features for the generated polygons are calculated (section 0). The dataset containing the features (X-values) and the targets (y-values) is then supplied to regression models for training. Finally, the prediction performance is evaluated using (untouched) validation test sets.

Fig. 1
Flowchart describing the proposed predictive modeling design pipeline
Fig. 1
Flowchart describing the proposed predictive modeling design pipeline
Close modal

3.1 Wavelength, Dimensions, and Materials.

For the current study, we consider radiation of wavelengths between 400 nm and 700 nm, a narrowband range that encompasses the peak of the solar irradiance. The studied nanowire cross sections are characterized with polygons of dimensions in the same order as the wavelength. We therefore consider the polygon radius to be less than 50 nm. For data generation, object materials are supposed to be silver (Ag).

4 Construction of Random Shapes

The arbitrary shapes are represented by polygons due to their geometric diversity and ease of construction. A polygon is a closed shape in R2 that can be fully characterized via its adjacent vertex coordinates. For a shape to be a valid polygon, the segments connecting adjacent vertices must not intersect. Our random polygon generation involves the following steps:

  1. Randomly select the number of vertices, Nv between 3 and Nvmax=10.

  2. Divide the coordinate system to Nv angular segments (to avoid intersection),

  3. In each segment, randomly select a radius between (5 nm–50 nm) and an angular position for each vertex.

This procedure is very similar to that of Ref. [46]. Figure 2 in Ref. [46] also provides some educational visuals that the reader may find interesting. The Cartesian and radial coordinates of the ith vertex are represented by Vi and Ψi. All polygons in our database are represented with equisized vectors. To do so, we use a different mathematical representation, where the polygon boundary is divided into nd = 360 elements in this study. Figure 2 shows the schematic of a polygon cross section of a hypothetical nanowire. i is the index of a vertex and Nv is the number of vertices in the given polygon. j is the index of boundary points. The point G indicates the centroid of a polygon, the location of which is calculated as
G=[gxgy]=13i=1Nv(xiyi+1xi+1yi)×[j=1nd(xj+xj+1)(xjyj+1xj+1yj)j=1nd(yj+yj+1)(xjyj+1xj+1yj)]
(4)
where nd is the number of boundary points.
Fig. 2
Schematic of a polygon cross section of a nanowire: left: three-dimensional polygon-prism with the schematics of the monitors used in FDTD calculations and right: definition of vertices, boundary points, and centroid
Fig. 2
Schematic of a polygon cross section of a nanowire: left: three-dimensional polygon-prism with the schematics of the monitors used in FDTD calculations and right: definition of vertices, boundary points, and centroid
Close modal

5 Predictive Geometric Features

Arbitrary shapes can be characterized using techniques adopted from object recognition where predictive features are used in classification. The available methods differ in whether they use the object boundary (boundary-based methods) or the interior points of the shapes (region-based methods) [60]. Since the polygons in this study have no holes, the boundary-based methods seem more appropriate. Some of the most widely used approaches for shape representation are polygonal approximations, interrelation evaluation, moments, transforms, chain codes [61], beam angle statistics [62], shape context [63], chord distributions [64], Fourier and wavelet transforms of shape signatures [65,66], and other techniques [67]. Moments can be realized in various ways such as invariant moments [68,69], Zernike moments [70], and more. One of the approaches to quantifying the geometric features is the bounding rectangles [71,72] that can help in calculating orientation and elongation related features. Additionally, simple descriptors, such as area, perimeter, compactness, eccentricity, and perimeter-area ratios, can also be included [73]. For the detailed classification of the shape representation methods and specific examples, interested readers are referred to the comprehensive reviews in the literature [60,67,7479].

Most of the aforementioned representation methods characterize a shape via a distribution, which is a vector with nd boundary points. To reduce the number of inputs, we only consider the scalar geometric descriptors among the mentioned methods. Furthermore, contrary to the field of object recognition, the scale of the shapes and their orientation with respect to the light direction are important discriminators in the estimation of radiative properties; Therefore, we must include hybrid features to characterize the directional and local properties of shapes vis-à-vis the light radiation. For example, the geometric cross section of a particle with respect to a ray of light depends on the view angle. The symmetry of the particles along the radiation axis and perpendicular to it can have meaningful and different predictive sentiments. In summary, we consider the following criteria in our feature engineering:

  1. Shape boundary is sufficient: the shapes in this study are represented by straight lines with no holes inside.

  2. Features are scalars: the descriptors expressed as distributions are not considered to limit the feature vector size.

  3. Orientation matters: In the current study, the orientation of shapes is important because the light direction is single direction. Two polygons that can be rotationally converted to one another are considered different shapes.

  4. Scale is important: Size certainly changes scattering response. The scattering of light by a polygon-shaped object is different from that of the same polygon but scaled up or down in size. Therefore, there must be features characterizing the scale of particles. On the other hand, the features with a primary purpose other than the scale must be scale-independent.

The complete set of features used in this study is listed in Tables 17.

Table 1

Simple features

NvNumber of vertices
pnNormalized perimeter
AArea
ξeEccentricity
ξcCompactness
ξrRectangularity
ξαCoverage angle from an observer far from the shape
ξvThe ratio of the visible vertices from an observer to the total number of vertices
NvNumber of vertices
pnNormalized perimeter
AArea
ξeEccentricity
ξcCompactness
ξrRectangularity
ξαCoverage angle from an observer far from the shape
ξvThe ratio of the visible vertices from an observer to the total number of vertices
Table 2

Moment-based features

ϕ11st moment invariant
ϕ22nd moment invariant
ϕ33rd moment invariant
ϕ11st moment invariant
ϕ22nd moment invariant
ϕ33rd moment invariant
Table 3

Elongation-related features

ϕe,11st order elongation
ϕe,44th order elongation
lyThe ratio of the perimeter of the bounding rectangle located along the x-axis to the actual shape perimeter.
τx,maxThe ratio of the shape extends on the y-axis to the maximum shape extended along any direction.
τy,maxThe ratio of the shape extends on the x-axis to the maximum shape extended along any direction.
τx,avgThe ratio of the shape extends on the y-axis to the average shape extended along any direction.
τy,avgThe ratio of the shape extends on the x-axis to the average shape extended along any direction.
ϕe,11st order elongation
ϕe,44th order elongation
lyThe ratio of the perimeter of the bounding rectangle located along the x-axis to the actual shape perimeter.
τx,maxThe ratio of the shape extends on the y-axis to the maximum shape extended along any direction.
τy,maxThe ratio of the shape extends on the x-axis to the maximum shape extended along any direction.
τx,avgThe ratio of the shape extends on the y-axis to the average shape extended along any direction.
τy,avgThe ratio of the shape extends on the x-axis to the average shape extended along any direction.
Table 4

Orientation-related features

γOrientation angle
ρ+xDirectional perimeter in the vicinity of +x direction
ρ+yDirectional perimeter in the vicinity of +y direction
ρxDirectional perimeter in the vicinity of −x direction
ρyDirectional perimeter in the vicinity of −y direction
γOrientation angle
ρ+xDirectional perimeter in the vicinity of +x direction
ρ+yDirectional perimeter in the vicinity of +y direction
ρxDirectional perimeter in the vicinity of −x direction
ρyDirectional perimeter in the vicinity of −y direction
Table 5

Sharpness-related features

s+xDirectional sharpness in the vicinity of +x direction
s+yDirectional sharpness in the vicinity of +y direction
sxDirectional sharpness in the vicinity of −x direction
syDirectional sharpness in the vicinity of −y direction
s+xDirectional sharpness in the vicinity of +x direction
s+yDirectional sharpness in the vicinity of +y direction
sxDirectional sharpness in the vicinity of −x direction
syDirectional sharpness in the vicinity of −y direction
Table 6

Convexity-related features

cConvexity of the overall shape
c+xDirectional convexity in the vicinity of +x direction
c+yDirectional convexity in the vicinity of +y direction
cxDirectional convexity in the vicinity of −x direction
cyDirectional convexity in the vicinity of −y direction
cConvexity of the overall shape
c+xDirectional convexity in the vicinity of +x direction
c+yDirectional convexity in the vicinity of +y direction
cxDirectional convexity in the vicinity of −x direction
cyDirectional convexity in the vicinity of −y direction
Table 7

Symmetry-related features

σxSymmetry around x direction
σySymmetry around y direction
σxSymmetry around x direction
σySymmetry around y direction

6 Dataset

We generated a dataset of 8230 data points each representing a random polygon cross section. By running FDTD simulations of the light scattering around the corresponding nanowires, we generated a corresponding dataset of scattering spectra. We computed independent variables (features) as explained previously, as well as the dependent variables (i.e., target) obtained from the computational simulations. We decided to use the following normalized transformed quantity as the regression target
Targetlog(Cscat/A)
(5)
where Cscat is the scattering cross section and A is the physical area. We refer to Cscat/A as the scattering coefficient. The logarithm of the scattering coefficient provides a normalized target distribution. The distributions of different target choices are compared in Fig. 3. The scatter plots of individual features with respect to the target are presented in Fig. 4. The absolute Pearson intercorrelation of the features is calculated and is graphically demonstrated after using a hierarchical clustering approach in Fig. 5. The heat map represents the 0-to-1 covariance matrix of the features while the tree chart on the left of the figure shows the distance-based hierarchical clustering.
Fig. 3
Histograms of different options for the target. The charts are aggregated over all randomized feature values (no value is fixed). Wavelength (λ) range is 400–700 nm.
Fig. 3
Histograms of different options for the target. The charts are aggregated over all randomized feature values (no value is fixed). Wavelength (λ) range is 400–700 nm.
Close modal
Fig. 4
Scatter plots of target with each feature
Fig. 4
Scatter plots of target with each feature
Close modal
Fig. 5
Analysis of the dataset using hierarchical clustering and absolute correlation matrix (light to dark: 0 to 1). The plot is generated for all randomized feature values (no value if fixed). Wavelength range is 400–700 nm.
Fig. 5
Analysis of the dataset using hierarchical clustering and absolute correlation matrix (light to dark: 0 to 1). The plot is generated for all randomized feature values (no value if fixed). Wavelength range is 400–700 nm.
Close modal

6.1 Remarks.

Note that the wavelength λ is the feature least correlated with everything else, which is expected. The decision to include wavelength as a feature is to reduce the computational cost of learning and to avoid having to setup a multi-output regression framework. The other feature at the outermost leave of the hierarchy is the orientation angle, γ. As shown in Fig. 4, γ has a V-shape relationship with the target, unlike other features. Although the number of vertices (Nv) is also expected to have a small correlation with the rest of the features, it relates to the 3rd moment invariant (ϕ3) in the cluster and has a small nonzero correlation with a few other features, such as convexity (c) and the ratio of the visible vertices from an observer to the total number of vertices (ξv). Convexity and 1st moment invariant (ϕ1) are also highly correlated. The features belong to one of two main clusters: The first cluster includes features that characterize local aspects of the shape, such as orientation, sharpness, convexity-related directional features, and the ratio of the shape extent in x and y directions to the average. The second cluster includes features that explain the more holistic characteristics of the shape such as simple, moment-based, and sharpness related features.

7 Data Fitting

We use a variety of models with different degrees of simplicity to study data fitting and modeling. We choose linear regression as the simple models, and regression trees, Extreme gradient boosted trees (XGBoost), and neural networks as the more complex nonlinear regression models for data fitting. XGBoost combines weak learners with strong learners consequently reducing the chance of overfitting [80]. Linear models are generally good choices when high-quality features are available without the need for additional normalization or transformation. Otherwise, they will be considerably underfitting. Even in the case of quality features, linear models can quickly suffer from overfitting in the presence of feature collinearity. The nonlinear models on the other hand are mostly prone to overfitting and need careful regularization and training cross-validation.

The model performance is monitored during the training using a validation set, and training is terminated if the validation set error starts increasing. We also deploy a ten-fold cross-validation scheme to alternate the validation set within the overall set and then average the model parameters to reduce any potential data bias. The model hyperparameters (e.g., size and configuration of the neural network, depth of the tree, etc.) are determined using a grid-search hyperparameter optimization with separate cross-validation. The ultimate predictor performance of every regression model is then assessed using an untouched test set. The data modeling and fitting tasks of the current work were implemented in python (v. 3.6) using Sklearn (v. 0.22) and XGBoost package (v. 1.3.0) [80,81].

8 Results and Discussion

8.1 Training Results.

In Table 8, the summary of the training results is presented. The training and validation set errors are shown with the coefficients of determination, or R-squared (R2). Figure 6 shows the scatter plots of the target and predicted outputs for each predictor. The linear regression algorithm has the highest error and lowest R2, as expected since the input-output relationship is highly nonlinear. Regression trees show an R2 larger than 0.85, yet it is significantly smaller than those of XGBoost and neural networks. One of the possible reasons for the poorer performance of regression trees is overfitting. As the tree structure gets complex, the generalization capability of the regression trees reduces.4

Fig. 6
Comparison of the target and predicted outputs by (a) regression trees, (b) XGBoost regression, (c) neural networks, and (d) linear regression. Left: training set, right: validation set.
Fig. 6
Comparison of the target and predicted outputs by (a) regression trees, (b) XGBoost regression, (c) neural networks, and (d) linear regression. Left: training set, right: validation set.
Close modal
Table 8

Summary of the performances of fitting methods.

TrainingValidation
MethodMSER2MSER2
Regression trees0.005580.9970.3558550.866
XGBoost regression4.07 × 10−71.0000.1270110.952
Neural networks0.0758950.9720.127480.952
Linear regression1.2472620.5391.3123030.506
TrainingValidation
MethodMSER2MSER2
Regression trees0.005580.9970.3558550.866
XGBoost regression4.07 × 10−71.0000.1270110.952
Neural networks0.0758950.9720.127480.952
Linear regression1.2472620.5391.3123030.506

Note: MSE: Mean-squared-error.

XGBoost and neural networks perform the best for predicting the output according to the validation error. Although the smaller training error of XGBoost compared to neural networks suggests overfitting, the inherent regularization of XGBoost results in a small validation error. On the other hand, neural networks also perform similarly on the validation set and the larger training error is due to the Bayesian regularization scheme employed during neural network training. As a result of the training, XGBoost and neural networks can be considered the best predictors.

8.2 Test Results.

The test dataset cases are entirely new to the predictors; thus, the test results show the real-world performance of the predictors. The previous section shows that XGBoost and neural networks are chosen as the best performing predictors and used for the test cases in this section. Figure 7 compares the target and the predictions obtained by XGBoost (Fig. 7(a)) and neural networks (Fig. 7(b)). The error rates and R2 values of these predictors are very close, which is in agreement with the training and validation results. The test R2 results are clearly slightly decaying compared to the validation cases. The performance of the final predictive models can be further improved by employing adaptive sampling techniques in which new training points are added from the high-error regions, possibly through an adversarial data-generation/training framework.

Fig. 7
Comparison of target and predicted data for the test set: (a) XGBoost and (b) neural networks regressions
Fig. 7
Comparison of target and predicted data for the test set: (a) XGBoost and (b) neural networks regressions
Close modal

In Fig. 8, comparison of the scattering cross section, Cs over the wavelength range in consideration is shown for three different cases. These examples are kept at a limited number for the sake of brevity. Note that in these plots, the target, log scattering efficiency (log Cs/A), is converted to the scattering cross section, Cs, for visualization purposes.

Fig. 8
Examples of the comparison of the target and predictions from XGBoost and neural networks over the wavelength spectrum
Fig. 8
Examples of the comparison of the target and predictions from XGBoost and neural networks over the wavelength spectrum
Close modal

The discrepancy between the target and the predictions is more visible at the peak locations of the scattering. It is expected to see discrepancy at large scattering values because data with large output is slightly less than the rest of the data. One possible approach to improve the fit at the peak locations is to create a supplementary predictor for predicting peak locations and peak values of each case. This approach can work as a refinement step of the prediction.

9 Conclusion

In this study, the optical characteristics of nanowires with arbitrary shape cross sections were modeled using data-driven techniques. The arbitrary shapes (randomly generated polygons) were characterized by the predictive shape descriptors, describing unique features of each shape. These features include simple properties such as area and eccentricity and more complex features presenting sharpness, convexity, and other more advanced geometric properties. These predictive shape features were used in regression frameworks to estimate optical scattering, namely the log scattering efficiency. Among different techniques, XGBoost and neural networks regressions proved to result in the smallest validation error. The test set experiments demonstrated that the predictors perform well out-sample, i.e., with respect to entirely new (test) cases. The final analytical predictive models are the XGBoost and NN feed forward models.

Although there are discrepancies between the target and the predictions, the general trend of the scattering was predicted very closely with the devised model. The peak (spectral) locations are estimated closely even though there is some mismatch in the peak value in some cases. The performance of the final predictive models can be further improved by employing adaptive sampling techniques in which new training points are added from the high-error regions, possibly through an adversarial data generation/training framework. This shall be a subject of future study.

Footnotes

2

This is assuming that the studied shape is sufficiently smooth, i.e., does not have sporadic significant spatial variations, which is a reasonable assumption.

3

This is named after Kane Yee, who developed the FDTD method.

4

This can potentially be alleviated with random forests, but XGboost is generally known to outperform that.

Acknowledgment

This work was supported by the US National Science Foundation (Grant No. CBET-2103008).

Conflict of Interest

There are no conflicts of interest.

Data Availability Statement

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.

Nomenclature

=

optical index (a, e, or s for absorption, extinction, or scattering, respectively)

A =

area

C =

optical cross section

G =

polygon centroid vector

E =

electric field

H =

magnetic field

ri =

polar coordinate radius

pn =

normalized perimeter

nd =

number of boundary points

Cs =

scattering cross section

I0 =

irradiance

Nv =

number of polygon vertices

Pj =

boundary point

Vi =

vertex point

gx, gy =

coordinates of the polygon centroid

xi, yi =

point coordinates

Pa, Pe, Ps =

power (absorbed, extincted, and scattered)

Greek Symbols

θi =

polar angle

Ψi =

polar coordinate vector

References

1.
Amendola
,
V.
,
Pilot
,
R.
,
Frasconi
,
M.
,
Maragò
,
O. M.
, and
Iatì
,
M. A.
,
2017
, “
Surface Plasmon Resonance in Gold Nanoparticles: A Review
,”
J. Phys.: Condens. Matter
,
29
(
20
), p.
203002
.
2.
Abadeer
,
N. S.
, and
Murphy
,
C. J.
,
2016
, “
Recent Progress in Cancer Thermal Therapy Using Gold Nanoparticles
,”
J. Phys. Chem. C
,
120
(
9
), pp.
4691
4716
.
3.
Tomchuk
,
P. M.
, and
Butenko
,
D. V.
,
2012
, “
The Nanoparticle Shape’s Effect on the Light Scattering Cross-Section
,”
Surf. Sci.
,
606
(
23–24
), pp.
1892
1898
.
4.
Ferry
,
V. E.
,
Polman
,
A.
, and
Atwater
,
H. A.
,
2011
, “
Modeling Light Trapping in Nanostructured Solar Cells
,”
ACS Nano
,
5
(
12
), pp.
10055
10064
.
5.
Montgomery
,
J. M.
,
Lee
,
T.-W.
, and
Gray
,
S. K.
,
2008
, “
Theory and Modeling of Light Interactions With Metallic Nanostructures
,”
J. Phys.: Condens. Matter
,
20
(
32
), p.
323201
.
6.
Kelly
,
K. L.
,
Coronado
,
E.
,
Zhao
,
L. L.
, and
Schatz
,
G. C.
,
2003
, “
The Optical Properties of Metal Nanoparticles: The Influence of Size, Shape, and Dielectric Environment
,”
J. Phys. Chem. B
,
107
(
3
), pp.
668
677
.
7.
Nehl
,
C. L.
, and
Hafner
,
J. H.
,
2008
, “
Shape-Dependent Plasmon Resonances of Gold Nanoparticles
,”
J. Mater. Chem.
,
18
(
21
), p.
2415
.
8.
Jain
,
P. K.
, and
El-Sayed
,
M. A.
,
2010
, “
Plasmonic Coupling in Noble Metal Nanostructures
,”
Chem. Phys. Lett.
,
487
(
4–6
), pp.
153
164
.
9.
Barnes
,
W. L.
,
2016
, “
Particle Plasmons: Why Shape Matters
,”
Am. J. Phys.
,
84
(
8
), pp.
593
601
.
10.
Yu
,
S.
,
Zhang
,
Y.
,
Wang
,
C.
,
Lee
,
W.
,
Dong
,
B.
,
Odom
,
T. W.
,
Sun
,
C.
, and
Chen
,
W.
,
2017
, “
Characterization and Design of Functional Quasi-Random Nanostructured Materials Using Spectral Density Function
,”
ASME J. Mech. Des.
,
139
(
7
), p.
071401
. doi.org/10.1115/1.4036582
11.
Fan
,
X.
,
Zheng
,
W.
, and
Singh
,
D. J.
,
2014
, “
Light Scattering and Surface Plasmons on Small Spherical Particles
,”
Light Sci. Appl.
,
3
(
6
), p.
e179
.
12.
Martin
,
R. J.
,
1993
, “
Mie Scattering Formulae for Non-Spherical Particles
,”
J. Mod. Opt.
,
40
(
12
), pp.
2467
2494
.
13.
Mishchenko
,
M. I.
,
Hovenier
,
J. W.
, and
Travis
,
L. D.
,
2000
,
Light Scattering by Nonspherical Particles: Theory, Measurements, and Applications
,
Academic Press
,
San Diego, CA
.
14.
Ren
,
Y.
,
Qi
,
H.
,
Chen
,
Q.
,
Wang
,
S.
, and
Ruan
,
L.
,
2017
, “
Localized Surface Plasmon Resonance of Nanotriangle Dimers at Different Relative Positions
,”
J. Quant. Spectrosc. Radiat. Transf.
,
199
, pp.
45
51
.
15.
Roopak
,
S.
,
Ji
,
A.
,
Parashar
,
P. K.
, and
Sharma
,
R. P.
,
2017
, “
Light Incoupling Tolerance of Resonant and Nonresonant Metal Nanostructures Embedded in Perovskite Medium: Effect of Various Geometries on Broad Spectral Resonance
,”
J. Phys. D: Appl. Phys.
,
50
(
33
), p.
335105
.
16.
Jain
,
P. K.
,
Lee
,
K. S.
,
El-Sayed
,
I. H.
, and
El-Sayed
,
M. A.
,
2006
, “
Calculated Absorption and Scattering Properties of Gold Nanoparticles of Different Size, Shape, and Composition: Applications in Biological Imaging and Biomedicine
,”
J. Phys. Chem. B
,
110
(
14
), pp.
7238
7248
.
17.
Agrawal
,
A.
,
Kriegel
,
I.
, and
Milliron
,
D. J.
,
2015
, “
Shape-Dependent Field Enhancement and Plasmon Resonance of Oxide Nanocrystals
,”
J. Phys. Chem. C
,
119
(
11
), pp.
6227
6238
.
18.
Raziman
,
T. V.
, and
Martin
,
O. J. F.
,
2013
, “
Polarisation Charges and Scattering Behaviour of Realistically Rounded Plasmonic Nanostructures
,”
Opt. Express
,
21
(
18
), p.
21500
.
19.
Sosa
,
O.
,
Noguez
,
C.
, and
Barrera
,
R. G.
,
2003
, “
Optical Properties of Metal Nanoparticles With Arbitrary Shapes
,”
J. Phys. Chem. B
,
107
(
26
), pp.
6269
6275
.
20.
Lombardi
,
A.
,
Loumaigne
,
M.
,
Crut
,
A.
,
Maioli
,
P.
,
Del Fatti
,
N.
,
Vallée
,
F.
,
Spuch-Calvar
,
M.
,
Burgin
,
J.
,
Majimel
,
J.
, and
Tréguer-Delapierre
,
M.
,
2012
, “
Surface Plasmon Resonance Properties of Single Elongated Nano-Objects: Gold Nanobipyramids and Nanorods
,”
Langmuir
,
28
(
24
), pp.
9027
9033
.
21.
Giannini
,
V.
,
Rodríguez-Oliveros
,
R.
, and
Sánchez-Gil
,
J. A.
,
2010
, “
Surface Plasmon Resonances of Metallic Nanostars/Nanoflowers for Surface-Enhanced Raman Scattering
,”
Plasmonics
,
5
(
1
), pp.
99
104
.
22.
Kwaadgras
,
B. W.
,
Verdult
,
M.
,
Dijkstra
,
M.
, and
van Roij
,
R.
,
2011
, “
Polarizability and Alignment of Dielectric Nanoparticles in an External Electric Field: Bowls, Dumbbells, and Cuboids
,”
J. Chem. Phys.
,
135
(
13
), p.
134105
.
23.
Molesky
,
S.
,
Lin
,
Z.
,
Piggott
,
A. Y.
,
Jin
,
W.
,
Vucković
,
J.
, and
Rodriguez
,
A. W.
,
2018
, “
Inverse Design in Nanophotonics
,”
Nat. Photon
,
12
(
11
), pp.
659
670
.
24.
Campbell
,
S. D.
,
Sell
,
D.
,
Jenkins
,
R. P.
,
Whiting
,
E. B.
,
Fan
,
J. A.
, and
Werner
,
D. H.
,
2019
, “
Review of Numerical Optimization Techniques for Meta-Device Design [Invited]
,”
Opt. Mater. Express
,
9
(
4
), p.
1842
.
25.
Wang
,
D.
, and
Su
,
G.
,
2014
, “
New Strategy to Promote Conversion Efficiency Using High-Index Nanostructures in Thin-Film Solar Cells
,”
Sci. Rep.
,
4
(
1
), pp.
1
7
.
26.
Wang
,
C.
,
Yu
,
S.
,
Chen
,
W.
, and
Sun
,
C.
,
2013
, “
Highly Efficient Light-Trapping Structure Design Inspired by Natural Evolution
,”
Sci Rep
,
3
(
1
), p.
1025
.
27.
Yu
,
S.
,
Wang
,
C.
,
Sun
,
C.
, and
Chen
,
W.
,
2014
, “
Topology Optimization for Light-Trapping Structure in Solar Cells
,”
Struct. Multidisc. Optim.
,
50
(
3
), pp.
367
382
.
28.
Dühring
,
M. B.
, and
Sigmund
,
O.
,
2013
, “
Optimization of Extraordinary Optical Absorption in Plasmonic and Dielectric Structures
,”
J. Opt. Soc. Am. B
,
30
(
5
), p.
1154
.
29.
Madsen
,
S. P.
,
Christiansen
,
J.
,
Christiansen
,
R. E.
,
Vester-Petersen
,
J.
,
Møller
,
S. H.
,
Lakhotiya
,
H.
,
Nazir
,
A.
, et al
,
2020
, “
Improving the Efficiency of Upconversion by Light Concentration Using Nanoparticle Design
,”
J. Phys. D: Appl. Phys.
,
53
(
7
), p.
073001
.
30.
Balling
,
P.
,
Christiansen
,
J.
,
Christiansen
,
R. E.
,
Eriksen
,
E.
,
Lakhotiya
,
H.
,
Mirsafaei
,
M.
,
Møller
,
S. H.
, et al
,
2018
, “
Improving the Efficiency of Solar Cells by Upconverting Sunlight Using Field Enhancement From Optimized Nano Structures
,”
Opt. Mater.
,
83
, pp.
279
289
.
31.
Wang
,
F.
,
Jensen
,
J. S.
, and
Sigmund
,
O.
,
2011
, “
Robust Topology Optimization of Photonic Crystal Waveguides With Tailored Dispersion Properties
,”
J. Opt. Soc. Am. B
,
28
(
3
), p.
387
.
32.
Burger
,
S.
,
Hammerschmidt
,
M.
,
Schneider
,
P.-I.
,
Garcia Santiago
,
X.
,
Zschiedrich
,
L.
, and
Weiser
,
M.
,
2018
, “
Solving Inverse Problems Appearing in Design and Metrology of Diffractive Optical Elements by Using Bayesian Optimization
,”
Computational Optics II
,
10694
, pp.
38
45
.
33.
Hegde
,
R. S.
,
2019
, “
Accelerating Optics Design Optimizations With Deep Learning
,”
Opt. Eng.
,
58
(
6
), p.
1
.
34.
Zhang
,
T.
,
Wang
,
J.
,
Liu
,
Q.
,
Zhou
,
J.
,
Dai
,
J.
,
Han
,
X.
,
Zhou
,
Y.
, and
Xu
,
K.
,
2019
, “
Efficient Spectrum Prediction and Inverse Design for Plasmonic Waveguide Systems Based on Artificial Neural Networks
,”
Photonics Res.
,
7
(
3
), p.
368
.
35.
Liu
,
D.
,
Tan
,
Y.
,
Khoram
,
E.
, and
Yu
,
Z.
,
2019
, “
Training Deep Neural Networks for the Inverse Design of Nanophotonic Structures
,”
Conference on Lasers and Electro-Optics
,
San Jose, CA
,
May 5–10
.
36.
Chen
,
Y.
,
Zhu
,
J.
,
Xie
,
Y.
,
Feng
,
N.
, and
Liu
,
Q. H.
,
2019
, “
Smart Inverse Design of Graphene-Based Photonic Metamaterials by an Adaptive Artificial Neural Network
,”
Nanoscale
,
11
(
19
), pp.
9749
9755
.
37.
Peurifoy
,
J.
,
Shen
,
Y.
,
Jing
,
L.
,
Yang
,
Y.
,
Cano-Renteria
,
F.
,
DeLacy
,
B. G.
,
Joannopoulos
,
J. D.
,
Tegmark
,
M.
, and
Soljačić
,
M.
,
2018
, “
Nanophotonic Particle Simulation and Inverse Design Using Artificial Neural Networks
,”
Sci. Adv.
,
4
(
6
), p.
eaar4206
.
38.
Chugh
,
S.
,
Gulistan
,
A.
,
Ghosh
,
S.
, and
Rahman
,
B. M. A.
,
2019
, “
Machine Learning Approach for Computing Optical Properties of a Photonic Crystal Fiber
,”
Opt. Express
,
27
(
25
), p.
36414
.
39.
Asano
,
T.
, and
Noda
,
S.
,
2018
, “
Optimization of Photonic Crystal Nanocavities Based on Deep Learning
,”
Opt. Express
,
26
(
25
), p.
32704
.
40.
Sajedian
,
I.
,
Badloe
,
T.
, and
Rho
,
J.
,
2019
, “
Optimisation of Colour Generation From Dielectric Nanostructures Using Reinforcement Learning
,”
Opt. Express
,
27
(
4
), p.
5874
.
41.
Nadell
,
C. C.
,
Huang
,
B.
,
Malof
,
J. M.
, and
Padilla
,
W. J.
,
2019
, “
Deep Learning for Accelerated All-Dielectric Metasurface Design
,”
Opt. Express
,
27
(
20
), p.
27523
.
42.
Alagappan
,
G.
, and
Png
,
C. E.
,
2019
, “
Modal Classification in Optical Waveguides Using Deep Learning
,”
J. Mod. Opt.
,
66
(
5
), pp.
557
561
.
43.
Otanicar
,
T.
,
Taylor
,
R. A.
,
Phelan
,
P. E.
, and
Prasher
,
R.
,
2009
, “
Impact of Size and Scattering Mode on the Optimal Solar Absorbing Nanofluid
,”
ASME 2009 3rd International Conference on Energy Sustainability, Volume 1
,
San Francisco, CA
,
Jan.
, pp.
791
796
.
44.
Malkiel
,
I.
,
Mrejen
,
M.
,
Nagler
,
A.
,
Arieli
,
U.
,
Wolf
,
L.
, and
Suchowski
,
H.
,
2018
, “
Plasmonic Nanostructure Design and Characterization Via Deep Learning
,”
Light Sci. Appl.
,
7
(
1
), p.
60
.
45.
Malkiel
,
I.
,
Mrejen
,
M.
,
Nagler
,
A.
,
Arieli
,
U.
,
Wolf
,
L.
, and
Suchowski
,
H.
,
2018
, “
Deep Learning for the Design of Nano-Photonic Structures
,”
IEEE International Conference on Computational Photography (ICCP)
,
Pittsburgh, PA
,
May
.
46.
Inampudi
,
S.
, and
Mosallaei
,
H.
,
2018
, “
Neural Network Based Design of Metagratings
,”
Appl. Phys. Lett.
,
112
(
24
), p.
241102
.
47.
Jiang
,
J.
,
Sell
,
D.
,
Hoyer
,
S.
,
Hickey
,
J.
,
Yang
,
J.
, and
Fan
,
J. A.
,
2019
, “
Free-Form Diffractive Metagrating Design Based on Generative Adversarial Networks
,”
ACS Nano
,
13
(
8
), pp.
8872
8878
.
48.
Liu
,
Z.
,
Zhu
,
D.
,
Rodrigues
,
S. P.
,
Lee
,
K.-T.
, and
Cai
,
W.
,
2018
, “
Generative Model for the Inverse Design of Metasurfaces
,”
Nano Lett.
,
18
(
10
), pp.
6570
6576
.
49.
Elzouka
,
M.
,
Yang
,
C.
,
Albert
,
A.
,
Prasher
,
R. S.
, and
Lubner
,
S. D.
,
2020
, “
Interpretable Forward and Inverse Design of Particle Spectral Emissivity Using Common Machine-Learning Models
,”
Cell Rep. Phys. Sci.
,
1
(
12
), p.
100259
.
50.
Garahan
,
A.
,
Pilon
,
L.
,
Yin
,
J.
, and
Saxena
,
I.
,
2006
, “
Optical Properties of Nanocomposite Thin-Films
,”
ASME International Mechanical Engineering Congress and Exposition
,
Chicago, IL
,
January
, Vol. 47845, pp.
177
186
.
51.
Mishchenko
,
M. I.
,
Travis
,
L. D.
, and
Lacis
,
A. A.
,
2002
,
Scattering, Absorption, and Emission of Light by Small Particles
,
Cambridge University Press
,
Cambridge, UK
.
52.
Hulst
,
H. C.
, and
van de Hulst
,
H. C.
,
1981
,
Light Scattering by Small Particles
,
Courier Corporation
,
North Chelmsford, MA
.
53.
Bohren
,
C. F.
, and
Huffman
,
D. R.
,
1983
,
Absorption and Scattering by Small Particles
,
John Wiley & Sons, Inc.
,
Hoboken, NJ
.
54.
Wang
,
W.
,
Chi
,
X.
,
Feng
,
Y.
, and
Zhao
,
Y.
,
2012
, “
Parallel FDTD Simulation of Photonic Crystals and Thin-Film Solar Cells
,”
13th International Conference on Parallel and Distributed Computing, Applications and Technologies
,
Beijing, China
,
Dec. 14–16
.
55.
Schneider
,
J.
,
2010
, “
Understanding the Finite-Difference Time-Domain Method
,” School of Electrical Engineering and Computer Science, Washington State University, p.
181
.
56.
Taflove
,
A.
,
1998
,
Computational Electrodynamics, Vol. 28
,
Artech House
,
Norwood, MA
.
57.
Fadakar Masouleh
,
F.
,
Das
,
N.
, and
Rozati
,
S.
,
2016
, “
Nano-Structured Gratings for Improved Light Absorption Efficiency in Solar Cells
,”
Energies
,
9
(
9
), p.
756
.
58.
Chung
,
H.
,
Jung
,
K.-Y.
,
Tee
,
X. T.
, and
Bermel
,
P.
,
2014
, “
Time Domain Simulation of Tandem Silicon Solar Cells With Optimal Textured Light Trapping Enabled by the Quadratic Complex Rational Function
,”
Opt. Express
,
22
(
S3
), p.
A818
.
59.
Mätzler
,
C.
,
2002
,
MATLAB Functions for Mie Scattering and Absorption
, Version 2.
60.
Loncaric
,
S.
,
1998
, “
A Survey of Shape Analysis Techniques
,”
Pattern Recognit.
,
31
(
8
), pp.
983
1001
.
61.
Karczmarek
,
P.
,
Kiersztyn
,
A.
,
Pedrycz
,
W.
, and
Dolecki
,
M.
,
2017
, “
An Application of Chain Code-Based Local Descriptor and Its Extension to Face Recognition
,”
Pattern Recognit.
,
65
, pp.
26
34
.
62.
Arica
,
N.
, and
Yarman Vural
,
F. T.
,
2003
, “
BAS: A Perceptual Shape Descriptor Based on the Beam Angle Statistics
,”
Pattern Recognit. Lett.
,
24
(
9–10
), pp.
1627
1639
.
63.
Belongie
,
S.
,
Malik
,
J.
, and
Puzicha
,
J.
,
2001
, “
Shape Context: A New Descriptor for Shape Matching and Object Recognition
,”
Adv. Neural Inf. Process. Syst.
,
13
, pp.
831
837
.
64.
Smith
,
S. P.
, and
Jain
,
A. K.
,
1982
, “
Chord Distributions for Shape Matching
,”
Comput. Graph. Image Process.
,
20
(
3
), pp.
259
271
.
65.
Andrews
,
M.
, and
Kakarala
,
R.
,
2015
, “
Shape Simplification Through Polygonal Approximation in the Fourier Domain
,”
Intelligent Robots and Computer Vision XXXII: Algorithms and Techniques
,
9406
, pp.
91
99
.
66.
Yuan
,
Z.
,
Li
,
F.
,
Zhang
,
P.
, and
Chen
,
B.
,
2014
, “
Description of Shape Characteristics Through Fourier and Wavelet Analysis
,”
Chin. J. Aeronaut.
,
27
(
1
), pp.
160
168
.
67.
Yang
,
M.
,
Kpalma
,
K.
, and
Ronsin
,
J.
,
2008
, “
A Survey of Shape Feature Extraction Techniques
,”
Pattern Recognit.
,
15
(
7
), pp.
43
90
.
68.
Taubin
,
G.
, and
Cooper
,
D. B.
,
1991
, “
Recognition and Positioning of Rigid Objects Using Algebraic Moment Invariants
,”
Geometric Methods in Computer Vision, SPIE
,
1570
, pp.
175
186
.
69.
Dai
,
X.
, and
Khorram
,
S.
,
1999
, “
A Feature-Based Image Registration Algorithm Using Improved Chain-Code Representation Combined With Invariant Moments
,”
IEEE Trans. Geosci. Remote Sens.
,
37
(
5
), pp.
2351
2362
.
70.
Zhao
,
Y.
,
Wang
,
S.
,
Zhang
,
X.
, and
Yao
,
H.
,
2013
, “
Robust Hashing for Image Authentication Using Zernike Moments and Local Features
,”
IEEE Trans. Inf. Forensics Secur.
,
8
(
1
), pp.
55
63
.
71.
Zunic
,
J.
,
Rosin
,
P. L.
, and
Kopanja
,
L.
,
2006
, “
On the Orientability of Shapes
,”
IEEE Trans. Image Process.
,
15
(
11
), pp.
3478
3487
.
72.
Rosin
,
P. L.
, and
Žunić
,
J.
,
2008
, “2D Shape Measures for Computer Vision,”
Handbook of Applied Algorithms: Solving Scientific, Engineering and Practical Problems
,
A.
Nayak
, and
I.
Stojmenović
, eds.,
John Wiley & Sons, Inc
,
Hoboken, NJ
, pp.
347
371
.
73.
Peura
,
M.
, and
Iivarinen
,
J.
,
1997
, “
Efficiency of Simple Shape Descriptors
,”
Proceedings of the Third International Workshop on Visual Form.
,
Capri, Italy
,
May
, Vol. 5. pp.
443
451
.
74.
Zhang
,
D.
, and
Lu
,
G.
,
2004
, “
Review of Shape Representation and Description Techniques
,”
Pattern Recognit.
,
37
(
1
), pp.
1
19
.
75.
Patel
,
M. N.
, and
Tandel
,
P.
,
2016
, “
A Survey on Feature Extraction Techniques for Shape Based Object Recognition
,”
Int. J. Comput. Appl.
,
137
(
6
), pp.
16
20
.
76.
Kurnianggoro
,
L.
, and
Wahyono
,
K. H. J.
,
2018
, “
A Survey of 2D Shape Representation: Methods, Evaluations, and Future Research Directions
,”
Neurocomputing
,
300
, pp.
1
16
.
77.
Hu
,
M.-K.
,
1962
, “
Visual Pattern Recognition by Moment Invariants
,”
IRE Trans. Inf. Theory
,
8
(
2
), pp.
179
187
.
78.
Chen
,
C.-C.
,
1993
, “
Improved Moment Invariants for Shape Discrimination
,”
Pattern Recognit.
,
26
(
5
), pp.
683
686
.
79.
Tsai
,
W. W.-H.
, and
Chou
,
S.-L.
,
1991
, “
Detection of Generalized Principal Axes in Rotationally Symmetric Shapes
,”
Pattern Recognit.
,
24
(
2
), pp.
95
104
.
80.
Chen
,
T.
, and
Guestrin
,
C.
,
2016
, “
XGBoost: A Scalable Tree Boosting System
,”
Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
,
San Francisco, CA
,
August
, pp.
785
794
.
81.
Pedregosa
,
F.
,
Varoquaux
,
G.
,
Gramfort
,
A.
,
Michel
,
V.
,
Thirion
,
B.
,
Grisel
,
O.
,
Blondel
,
M.
, et al
,
2011
, “
Scikit-Learn: Machine Learning in Python
,”
J. Mach. Learn. Res.
,
12
, pp.
2825
2830
.