Abstract

     A major problem with deep generative models is verifying that the generated distribution resembles the target distribution while the individual generated sample is indistinguishable from the original data. In particular, for application in astrophysics we need to be sure that the generated data matches our prior knowledge and that the generated samples entail all object types with the correct frequency and diversity. We currently lack objective ways to systematically assess these quality aspects, where human inspection reaches its limits, as this requires detailed analysis of a large data volume. In this work, we identify reasonable metrics for the quality of galaxy image generators. To this end, we compare a small set of conditional image generators, trained on galaxy images with classification labels for visual morphology features. Our main contribution is a new set of cluster-based metrics for matching the generated distribution to the target distribution. Furthermore, we use the Wasserstein distance on proxies for galaxy morphology as well as a number of other metrics commonly used for image generators. The newly introduced cluster-based metrics are good proxies for the quality of the generated distribution and are suited for automatized identification of mode collapse. Furthermore, the cluster metrics allow for a qualitative interpretation of the generated distribution. The metrics based on morphological statistics provide a useful tool to probe the physical soundness of generated samples. Finally, we find that kernel inception distance used with an InceptionV3 model pre-trained on ImageNet is a good proxy for the overall quality of galaxy image generators, although it cannot be interpreted that easily.

Introduction

     Upcoming astronomical surveys with telescopes, such as Euclid (Laureijs et al., 2012) and LSST (Abell et al., 2009), will provide a wealth of data with billions of galaxy images. These hold unprecedented insight in highly researched astrophysical and cosmological questions, such as formation and evolution of galaxies, the cosmic distribution of dark matter, as well as the expansion history of the Universe (Laureijs et al., 2011). However, the huge amount of images is far too big to be investigated by astrophysicists on an individual basis. Instead, fast and systematic extraction of galaxy properties from their images is required in order to inform and constrain physical models. There are a number of computational tools that have been developed and are already in use by the astrophysical community (Rodriguez-Gomez et al., 2019, Boquien et al., 2019, Shamir, 2011). Though these allow for a systematic extraction of properties, they require too much computational resources to provide big collections of high quality mock images in a reasonable amount of time. Hence, it is necessary to replace these tools by faster methods, e. g. by using machine learning techniques, such as deep neural networks (Lovell et al., 2019, Ferreira et al., 2020, Walmsley et al., 2022 e. g.).

Conclusions

     In order to find suitable evaluation metrics to assess the quality of galaxy image generators, we investigate a number of evaluation metrics with the potential to measure different quality aspects. These metrics are probed with a small number of conditional generative models, some of which purposely are of worse quality than the others. We identified evaluation metrics that are good proxies for the quality of individual images and the resemblance of the target distribution. Our main results are:

• The newly introduced cluster-based metrics (Section 3.1) are a formidable new tool to assess the distribution of generated data.

• The KID metric on features of the pre-trained InceptionV3 model (Szegedy et al., 2015) provides a useful proxy on the overall quality of galaxy image generators.

• Using an ALReLU activation (Mastromichalakis, 2020) significantly enhances classification accuracy for rare object types. This is required to train conditional generators on the highly imbalanced dataset of galaxy images.

     We introduce a new set of cluster-based metrics (Section 3.1), which are good proxies for the distribution quality. In particular, they allow for a qualitative interpretation regarding the sample diversity and the amount of samples generated for different types. Thus, they provide an intuitive new tool to assess the resemblance of the target distribution in a qualitative way, which so far has been a rather illusive task. In addition, one of the cluster metrics, the cluster error provides a formidable tool to identify mode collapse. Moreover, the cluster metrics have the potential to be transformed into a loss function to train generative models to reproduce the distribution of training data exactly, or any other target distribution, e. g. balanced datasets. Together, the aforementioned metrics provide a good basis to assess the quality of trained image generators, which is not necessarily limited to the context of galaxy images. We are currently working on providing losses based on these cluster metrics, which have the potential to drastically improve training of generative models regarding distribution quality, which is especially valuable for GANs.

Highlights

Abstract

Keywords

1. Introduction

2. The five problems of irreversibility

3. Dynamical density functional theory

4. First problem: The source of irreversibility in thermodynamics

5. Second Problem: The Definition of “Equilibrium” and “Entropy”

6. Third problem: The justification of coarse-graining

7. Fourth problem: The (irreversible) approach to equilibrium

8. Fifth problem: The arrow of time

9. Conclusion

Declaration of Competing Interest

Acknowledgments

References

Abstract

Thermodynamics has a clear arrow of time, characterized by the irreversible approach to equilibrium. This stands in contrast to the laws of microscopic theories, which are invariant under time-reversal. Foundational discussions of this “problem of irreversibility” often focus on historical considerations, and do therefore not take results of modern physical research on this topic into account. In this article, I will close this gap by studying the implications of dynamical density functional theory (DDFT), a central method of modern nonequilibrium statistical mechanics not previously considered in philosophy of physics, for this debate. For this purpose, the philosophical discussion of irreversibility is structured into five problems, concerned with the source of irreversibility in thermodynamics, the definition of equilibrium and entropy, the justification of coarse-graining, the approach to equilibrium and the arrow of time. For each of these problems, it is shown that DDFT provides novel insights that are of importance for both physicists and philosophers of physics.

1. Introduction

The temporal asymmetry of thermodynamics is one of the central problems in philosophy of physics. If a cup of hot coffee stands in a room, it will cool down until it has room temperature, but it will not spontaneously heat up by extracting heat from its environment. This is commonly seen as a consequence of the second law of thermodynamics, which assigns to each of these systems a quantity known as “entropy” that increases in these processes and that, most importantly, cannot decrease. Often, this is considered one of the most fundamental laws of physics (Callender, 2001, p. 540).

However, as is also well-known, there is a conflict with the microscopic laws governing the motion of the individual particles that a macroscopic system consists of. These laws are invariant under timereversal,1 which means that if a process can occur in one direction of time, it can also occur in the other direction. Thus, a cup of coffee at room temperature that spontaneously heats up would be in perfect agreement with the microscopic laws of physics, which makes it very difficult to explain why such a behavior is never observed.

An intense discussion on this problem has emerged in philosophy of physics. It has evolved into a variety of sub-debates concerned with different explananda that, in this article will be classified into five problems. Often, foundational discussions of statistical mechanics focus on historical aspects. Moreover, there has been a growth of interest in formal approaches to coarse-graining based on the projection operator formalism (Mori, 1965; Zwanzig, 1960). There is, however, a lack of work that considers the implications of modern research on nonequilibrium statistical mechanics for foundational problems (Wallace, 2015).

1- Introduction

2- Stochastic inflation with large N

3- Example model

4- Discussion and conclusions

References

Abstract

We propose a class of single-field, slow-roll inflation models in which a typical number of e-folds can be extremely large. The key point is to introduce a very shallow local minimum near the top of the potential in a hilltop inflation model. In particular, a typical number of e-folds is enhanced if classical behavior dominates around the local minimum such that the inflaton probability distribution is drifted to the local minimum as a whole. After the inflaton escapes from the local minimum due to the stochastic dynamics, the ordinary slow-roll inflation follows and it can generate the primordial density perturbation consistent with observation. Interestingly, our scenario inherits the advantages of the old and new inflation: the typical e-folds can be extremely large as in the old inflation, and slow-roll inflation naturally follows after the stochastic regime as in the new inflation. In our numerical example, the typical number of e-folds can be as large as 101010 , which is large enough for various light scalars such the QCD axion to reach the Bunch-Davies distribution.

Introduction

How long can inflation last? From the observational point of view, the total number of e-folds, N, must be larger than ∼ 50–60, which weakly depends on the inflation scale and thermal history after inflation. On the other hand, an extremely long duration of inflation is often required in certain scenarios. For instance, the relaxation model [1], the stochastic axion scenario [2–5], a quintessence model [6,7] demand the e-folding number of log10 N ∼ O(10), depending on model parameters.1 The purpose of this Letter is to provide a simple single-field, slow-roll inflation model whose typical number of e-folds is extremely large. There is a variety of inflation models which last very long. In the string/axion landscape [10–17], there are many local minima where the old inflation takes place and continues until the inflaton tunnels toward one of the adjacent local minima with a lower energy through bubble formation. In this case, the typical e-folding number can be exponentially large. However, one needs slow-roll inflation after the bubble formation to explain the observed cosmic microwave background (CMB) temperature/polarization anisotropies. For this, one may need another light scalar, in which case the two inflation scales are not related to each other, in general.

1- Introduction

2- Model and principle

3- Results and discussion

4- Conclusions

References

Abstract

As an indirect imaging technique, computational ghost imaging (GI) obtains the object information by calculating the intensity correlation between a series of computer-generated matrices and the corresponding bucket signals, which thereby offers a potential application in optical encryption. Here, we propose a new steganography scheme, called ghost steganography, based on the principle of computational GI. In our ghost steganography scheme, the bucket intensity signals of a secret image are concealed into the ones of a non-secret image by applying a non-conspicuous number integration process. To further increase the security, we introduce RSA cryptography to encode the integrated bucket signals after the steganography process. Simulation and experiment results fully demonstrate the feasibility of our optical ghost cryptography and steganography scheme. Our work paves a way to the application of GI in steganography and also enriches the knowledge of symmetric and asymmetric optical cryptography.

Introduction

Ghost imaging (GI), also known as correlated imaging, is an indirect imaging modality which obtains the object information from the intensity fluctuation correlation of two beams. One beam, called object beam, going through the object, is measured by a bucket detector. The other beam, called reference beam, interacting without the object, is detected by a spatially resolved camera. GI was firstly achieved with entangled photon pairs experimentally in 1995 [1], and later extended into the classical region with various thermal light sources [2–11]. Different from the two-detector GI, Shapiro proposed the computational GI in 2008 [12], which generated the active illumination patterns by using the spatial light modulator instead of passive measurements of reference beam. Similar to the single-pixel camera technique [13], computational GI can recover the object image with only a single-pixel detector [14,15], which largely simplifies the experimental setup. Meanwhile, the correlation between the computer-generated random matrix and the object beam intensity of computational GI offered a potential application in optical encryption [16-30]. To increase the efficiency and security of optical encryption based on GI [16], different methods are proposed and developed, including gray-scale and color encryption [17], multiple-image encryptions [18,26,29], specific phase masks schemes [22,25,27,30], XOR operation scheme [24].

1- Introduction

2- Theory and proposed scheme

3- Result and discussion

4- Conclusion

References

Abstract

This paper reveals an analytical perception of electro-optically tunable cascaded Kerr-lens using RbTiOPO4 (RTP) crystal where the advantageous properties of the crystal have been studied to mitigate the limitations of KTiOPO4 (KTP) and BaB2O4 (BBO) in intracavity high power laser generation. An external electric field application to the z-direction of RTP is introducing a phase mismatch condition leading to a nonlinear phase shift within propagating optical waves. Based on simulated endeavors, a maximum effective nonlinear refractive index of ± 1.4053 × 10−13 cm2 /W is calculated by developing a phase shift of ∓4.6 rad. In contrast, KTP and RTP both have almost similar optical properties but due to certain special features like higher damage threshold, less walk-off angle etc., RTP is firmly accepted for high power applications. Additionally, this article is augmented to analyze the self-focusing and defocusing phenomenon owing to the nonuniform distribution of radial intensity of the propagating Gaussian beam. To the best of author’s knowledge, the present analysis manifests the superior performance of RTP that may pave the way to devise a configurable lens analogous to Kerr-lens having an electro-optically tunable focal length.

Introduction

In the past several decades [1–5], cascaded second-order nonlinearity has played a vital role in the generation of cascaded Kerr lens mode-locking [6] that is nothing but a successive combination of two second harmonic generation (SHG) processes (χ(2):χ(2) development) that yields an intensity-dependent large phase mismatch in propagating fundamental waves. Accordingly, the nonlinear phase shift because of the non phasematched condition is also relative to the intensity of the incident beam that is analogous to Kerr medium (χ(3) nonlinearity) [6,7]. Moreover, the performance of this feature may enlarge the third-order nonlinearity of a medium having weak third-order coefficient [8] which is grander than the natural third-order nonlinearity [5] associated with the Kerr electro-optic effect as well. As a consequence of that, cascaded second-order nonlinearity equivalent to third-order nonlinearity [2] plays a significant role in several areas like transistor action [9], electro-optical detection [10], electro-optic switching [11], selffocusing and self-defocusing [9] etc. [12–15]. Although different techniques are emphasized to develop cascaded second-order nonlinearity using various types of crystals where problems like undesirable cavity loss, inaccuracy of incident beam angle, etc [5,16]. have occurred that deters a controllable desired output. In recent years [17,18], an analytical method has been introduced named “electro-optically tunable cascaded nonlinearity” in an exceedingly bulk anisotropic crystals like Potassium Titanyl Phosphate (KTiOPO4, KTP) and Beta Barium Borate (BaB2O4, BBO), etc.

1- Introduction

2- Calculation details

3- Results and discussion

4- Conclusion

References

Abstract

High spin-polarized materials are expected for the development of spintronic devices. In this work, we are aiming to find new spintronic materials in Mn-based binary, ternary and quaternary Heusler alloys by using the first-principles calculations. We investigated structures, electronic properties, magnetic, tetragonal distortion and mechanic properties of Mn3Ga, Mn2YGa (Y = V, Nb, Ta) and ScMnVGa. The results show that Mn3Ga, ScMnVGa type-I, and type-II have lower energy in antiferromagnetic (AFM) states than ferromagnetic (FM) and paramagnetic (NM). Mn2YGa (Y = V, Nb, Ta) compounds are more stable in FM states. The results of electronic and magnetic properties indicate that Mn3Ga is a gapless half-metallic antiferromagnet (Gapless HM-AFM). Mn2VGa and Mn2NbGa belong to gapless half-metallic ferrimagnets (Gapless HM-FE). Mn2TaGa is ferrimagnets. ScMnVGa type-I is half-metallic antiferromagnets (HM-AFM) and type-II belongs to spin gapless semiconductors with AFM (SGSs-AFM). Inspired by Ferromagnetic shape memory alloys (FAMAs), we performed tetragonal deformation of those compounds, and we predicted that Mn3Ga, Mn2YGa (Y = V, Nb, Ta) have possible martensitic transformations. Finally, the mechanical stability and elastic properties of above-mentioned in both cubic and tetragonal structures were discussed in detail.

Introduction

Spintronics is considered to be one of the most promising subjects in the information age because spintronic devices use not only the state of charge but electronic spin to transfer and store information, which speeds up information transmission and reduces energy losses. The findings of Giant Magnetoresistance (GMR), Colossal Magnetoresistance (CMR) and Tunnel Magnetoresistance (MGR) accelerate the development of spintronics. Despite great potential advantages, spintronics still faces some challenges, such as, generate high spin injectors. However, the discovery of half-metallic ferrimagnet has improved this problem, which is an ideal semiconductor spin-injection source due to it has 100% spin polarizability. Among different halfmetallic materials, Heusler alloys are prominent especially owing to high Curie temperature and flexible electronic structures [1,2]. Since De Groot et al. [3] discovered the half-metallic ferromagnets of NiMnSb, while they were calculating the energy band using the planewave method, this intermetallic materials with unusual electronic structures arouse people’s attention. In the big family of Heusler alloys, Co-based Heusler alloys [4–6] usually show magnetism, and the source of this property is direct interaction between Co-Co atoms. Cu-based Heusler alloys [7–9] are typical representative of non-magnetic elements show ferromagnetic after high-ordered, Pd-based Heusler alloys [10,11] which contain rareearth elements is superconducting, Ni2Mn-based Heusler alloys [12–17] have both ferromagnetic properties and thermoelastic martensitic transformation, which realizing magnetic-field-induced strain, shape memory effect, magnetoresistance effect and other application functions. At present, most research focuses on ternary Mn-based Heusler alloys, and they usually behave ferromagnetic or ferromagnetic [18–21]. Actually, besides half-metallic ferromagnets and ferrimagnets, half-metallic antiferromagnets also worthy of study. Not only it has 100% spin polarizability, but its total magnetic moment is zero. Thus, it means that when HM-AFMs are applied to spintronic devices, and those alloys will get the feature of the low-stray field and low energy loss. HM-AFMs were discovered in Heusler alloys [22] and perovskite compounds [23]. Besides, the martensitic transformation was found in some Mn-based Heusler alloys [24–26].

Nomenclature

1. Introduction

2. Experimental section

3. Result and discussion

4. Conclusions

Acknowledgement

References

Abstract

The chemical looping process on honeycomb reactor for solar syngas is experimentally studied in this work, which is the key reaction of the liquid sunshine production process. The honeycomb reactor realizes the integration of oxygen carrier and reaction chamber. NiO is placed in the reactor as oxygen carrier and methane is introduced as fuel gas. The results show that, with the development of process, the major reaction in the reactor gradually changed from methane complete oxidation to methane partial oxidation. During the process, the methane conversion and outlet syngas concentration is affected by the methane flow and fractional oxidation. Under the optimal operating conditions, the methane conversion can be maintained more than 95% and the concentration of outlet syngas can be around 90%. Compared with non-honeycomb fixed bed reactor, the methane conversion increases by more than 20 percent point and the concentration of outlet syngas increases by about 10 percent point. In addition, oxygen carrier in honeycomb reactor shows excellent cyclic stability in 30 times experiments.

Introduction

When the consequences of the overuse of fossil fuels come, are we prepared to deal with them? Since the first industrial revolution, fossil fuels have been the primary source of global energy. Behind the exponential progress powered by fossil fuels are the potential of energy crisis and the environmental deterioration [1,2]. Nowadays, annual carbon dioxide emissions are about twice as much as that can be absorbed by nature [3]. If we continue to rely on fossil fuels, the resulting climate change and air pollution will threaten our life [4]. Sunlight is the most abundant source of energy on earth, and it can provide about 885 million TWh of energy in a year, surpassing other kinds of renewable energy by hundreds of times [5]. But it is hard to use it as easily as you flip a switch due to the decentralization and discontinuity of solar [6]. If we want to capture, store sunlight and supply it as energy source, the key process is to convert it into a stable, storable, highenergy-density chemical fuel.

The liquid sunshine vision is increasingly appealing to researchers [7]. Liquid fuels are not difficult to transport and store, and it can be widely used with some improvements to existing infrastructures. Liquid sunlight is designed to convert sunlight into liquid fuels such as methanol. Methanol is an attractive candidate because of its wide application. There are no C-C bonds in methanol which can effectively reduce the greenhouse gas emission due to its low carbon to hydrogen ratio [7]. As the raw material of methanol production, the preparation of solar syngas is a key process in current researches [8–11].

1- Introduction

2- Methodology

3- Results

4- Discussion

5- Conclusion

References

Abstract

Weather radar can provide spatially explicit precipitation grids. However interference, ground clutter and various causes of attenuation introduce uncertainty into the result. Typically, rain gauge observations, recognized as a precise measure of precipitation at point locations, are used to adjust weather radar grids to obtain more accurate precipitation maps. This adjustment involves one or more of various geostatistic techniques. Yet, since gauges are sparsely located, a geostatistic approach is sometimes limited or even not applicable.
This work adopts an alternative to radar adjustment by merging location-based variables with rain grids from weather radar. Recognizing that location-based variables: elevation, slope, aspect and distance from the coast all affect precipitation, these are applied to the original weather radar grid to produce an altered precipitation distribution.
The merging procedure presented here uses fuzzy logic, whereby all variables, as well as the original radar are assigned probabilities known as membership functions (MF), then a joint membership function (JMF) combines all MFs in the fuzzy set, each multiplied by its weight, to create a precipitation probability grid. This JMF probability grid is validated with gauge observation data. We show up to 30% higher correlation coefficients between gauges and the JMF grid than between gauges and the original radar. The improved correlation results from the flexibility of fuzzy logic in transforming location-based variables to probabilities.

Background

Estimating spatially distributed precipitation grids is a prerequisite to flood management and flood forecasting (Merz et al. (2014)). Hydrological models need basin-scale, spatially explicit precipitation data, among other inputs, to construct accurate flood forecasts (Todini et al. (2005)) for surface runoff management. Rain radar can produce such spatial precipitation distributions, however the challenges in calibrating and correcting for the various sources of error (detailed in Villarini et al. (2008)) create spatial and temporal uncertainty in the precipitation distribution (Cecinati et al. (2017), Krajewski and Smith, 2002). Nevertheless, the underlying motivation for research in improving precipitation maps rests in the needs of hydrological modeling and flood forecasting. Since weather radar became an accepted source of spatially distributed rainfall (Krajewski and Smith (2002), Morin et al. (2003)), extensive research has examined adjustment procedures to merge rain gauge observations with weather radar. Gauge data are accepted as reference observations (see for example Colli et al. (2013)), but represent point locations. Such point data can adjust weather radar grids through several geostatistic methods, reviewed and evaluated by Goovaerts (2000), Berndt et al. (2014) and McKee and Binns (2016). Kriging based methods have been examined by Kebaili Bargaoui and Chebbi (2009), Adhikary et al. (2017), and Ly et al. (2013). A comparison of various kriging methods where elevation was the secondary variable was done by Carrera-Hernández and Gaskin (2007). Another unique algorithm known as Conditional Merging, developed and evaluated by Sinclair and Pegram (2005), applies multiple kriging steps to achieve successful adjustment (Kim et al. (2007)) of weather radar grids.

I- Introduction

II- Experiment

III- Analysis of Experimental Results

IV- Conclusion

References

Abstract

As one of the most important external insulation equipment in power system, insulators play an important role in transmission and distribution network. In this paper, qualitatively and quantitatively analysis of the contaminated elements on the surface of artificial contaminated insulators and natural contaminated insulators were carried out by Laser-Induced Breakdown Spectroscopy (LIBS). For artificial contaminated insulators, Na and Al elements can be used to characterize the equivalent salt deposit density (ESDD) and non-soluble deposit density (NSDD) levels respectively. For natural contaminated insulators, there are differences in spectral intensity of contaminated elements with different contamination years. And with the increase of the contamination years, the spectral intensity of the Si, Fe, Ca, Al, K and Na elements also have an increasing trend. Insulator samples under 4 contamination levels were classified based on the principal component analysis (PCA) algorithm. Dimension reduction analysis was conducted by extracting 6 excitation spectra (Si, Fe, Ca, Al, K, Na) which can represent contamination information. The 2 principal components with the largest contribution rate were obtained and the principal component scores of each spectrum were calculated. The research results showed that the spectral sample points have obvious convergence phenomenon according to the contamination level of the insulators. Compared with the commonly used methods for judging contamination level, LIBS technology has the advantages of easy operation, no sample processing and high sensitivity. Therefore, the combination of LIBS technology and PCA algorithm can save time and cost and improve the detection efficiency in the field of insulator.

INTRODUCTION

As one of the most important external insulation devices for power system, insulators play an important role in the transmission and distribution networks. With the development of China, the environmental pollution caused by the daily activities of industry and humans is also increasing. The contamination on insulator surfaces will affect its external insulation performance. In severe cases, it will cause flashover, and the flashover on transmission lines will cause large-scale power outage, which will result in power system failure and bring huge economic losses [1]. Therefore, the evaluation of insulator contamination is the basic research work in the field of external insulation of power systems, which is of great significance for the design, operation and mainten.ance of external insulation of power systems. A large number of scholars at home and abroad have conducted a lot of research on insulator contamination detection and achieved a series of outstanding results, which have greatly reducing the power system failure caused by contamination flashover [2,3]. At present, the most common method of measuring contamination is to remove the insulator from the power line and then determine the contamination level by measuring the equivalent salt deposit density and nonsoluble deposit density [4]. However, this method is complicated in operation and needs to be powered off, which is not convenient in practical operation [5]. Reference [6] installed metal electrodes on the surface of insulators, which can determine the contamination degree of insulators by measuring the surface conductivity of insulators when the air humidity is relatively saturated. The ultraviolet discharge pulse signal will appear when the leakage current is generated on the insulator surface. The relevant characteristics of the leakage current can be obtained and the contamination degree can be determined by analyzing the signal with a certain algorithm [7]. However, the exact relationship between the current leakage current and the degree of contamination has not been fully clarified, and the installation of the equipment is also complicated, and the degree of environmental impact is relatively large.