Searcharxiv⌕ Search

arXiv subjects

Pawan Kumar

Publications and source records attributed to Pawan Kumar.

At least 73 records · Page 4Linked to original sources
arXiv2024

A Gauss-Newton Approach for Min-Max Optimization in Generative Adversarial Networks

A novel first-order method is proposed for training generative adversarial networks (GANs). It modifies the Gauss-Newton method to approximate the min-max Hessian and uses the Sherman-Morrison inversion formula to calculate the inverse. The method corresponds to a fixed-point method that ensures necessary contraction. To evaluate its effectiveness, numerical experiments are conducted on various datasets commonly used in image generation tasks, such as MNIST, Fashion MNIST, CIFAR10, FFHQ, and LSUN. Our method is capable of generating high-fidelity images with greater diversity across multiple datasets. It also achieves the highest inception score for CIFAR10 among all compared methods, including state-of-the-art second-order methods. Additionally, its execution time is comparable to that of first-order min-max methods.

cs.LG↗
arXiv2024

Characterizing the solar cycle variability using nonlinear time series analysis at different amounts of dynamo supercriticality: Solar dynamo is not highly supercritical

The solar dynamo is essentially a cyclic process in which the toroidal component of the magnetic field is converted into the poloidal one and vice versa. This cyclic loop is disturbed by some nonlinear and stochastic processes mainly operating in the toroidal to poloidal part. Hence, the memory of the polar field decreases in every cycle. On the other hand, the dynamo efficiency and, thus, the supercriticality of the dynamo decreases with the Sun's age. Previous studies have shown that the memory of the polar magnetic field decreases with the increase of supercriticality of the dynamo. In this study, we employ popular techniques of time series analysis, namely, compute Higuchi's fractal dimension, Hurst exponent, and Multi-Fractal Detrended Fluctuation Analysis, to the amplitude of the solar magnetic cycle obtained from dynamo models operating at near-critical and supercritical regimes. We show that the magnetic field in the near-critical regime is governed by strong memory, less stochasticity, intermittency, and breakdown of self-similarity. On the contrary, the magnetic field in the supercritical region has less memory, strong stochasticity, and shows a good amount of self-similarity. Finally, applying the same time series analysis techniques in the reconstructed sunspot data of 85 cycles and comparing their results with that from models, we conclude that the solar dynamo is possibly operating near the critical regime and not too much supercritical regime. Thus Sun may not be too far from the critical dynamo transition.

astro-ph.SR↗
arXiv2024

Phonon-pair-driven Ferroelectricity Causes Costless Domain-walls and Bulk-boundary Duality

Ferroelectric domain walls, recognized as distinct from the bulk in terms of symmetry, structure, and electronic properties, host exotic phenomena including conductive walls, ferroelectric vortices, novel topologies, and negative capacitance. Contrary to conventional understanding, our study reveals that the structure of domain walls in HfO2 closely resembles its bulk. First, our first-principles simulations unveil that the robust ferroelectricity is supported by bosonic pairing of all the anionic phonons in bulk HfO2. Strikingly, the paired phonons strongly bond with each other and successfully reach the center of the domain wall without losing their integrity and produce bulk-like domain walls. We then confirmed preservation of the bulk phonon displacements and consequently full revival of the bulk structure at domain walls via aberration-corrected STEM. The newly found duality between the bulk and the domain wall sheds light on previously enigmatic properties such as zero-energy domain walls, perfect Ising-type polar ordering, and exceptionally robust ferroelectricity at the sub-nm scales. The phonon-pairing discovered here is robust against physical boundaries such as domain walls and enables zero momentum and zero-energy cost local ferroelectric switching. This phenomenon demonstrated in Si-compatible ferroelectrics provides a novel technological platform where data storage on domain walls is as feasible as that within the domains, thereby expanding the potential for high-density data storage and advanced ferroelectric applications.

cond-mat.mtrl-sci↗
arXiv2024

Motility and pair-wise interactions of chemically active droplets in 1-D confinement

Self-propelled droplets serve as ideal model systems to delve deeper into understanding of the motion of biological micro-swimmers by simulating their motility. Biological microorganisms are renowned for showcasing a diverse array of dynamic swimming behaviors when confronted with physical constraints. This study aims to elucidate the impact of physical constraints on swimming characteristics of biological microorganisms. To achieve this, we present observations on the individual and pair-wise behavior of micellar solubilized self-propelled 4-Cyano-4'-pentyl-biphenyl (5CB) oil droplets in a square capillary channel filled with a surfactant trimethyl ammonium bromide (TTAB) aqueous solution. To explore the effect of the underlying Péclet ($Pe$) number of the swimming droplets, the study is also performed in the presence of additives such as high molecular weight polymer Polyethylene oxide (PEO) and molecular solute glycerol. The capillary confinement restricts droplet to predominantly one-dimensional (1D) motion, albeit with noticeable differences in their motion across the three scenarios. Through a characterization of the chemical and hydrodynamic flow fields surrounding the droplets, we illustrate that the modification of the droplets' chemical field due to confinement varies significantly based on the underlying differences in the Péclet number ($Pe$) in these cases. This alteration in the chemical field distribution notably affects the individual droplets' motion. Moreover, these distinct chemical field interactions between the droplets also lead to variations in their pair-wise motion, ranging from behaviors like chasing to scattering.

cond-mat.soft↗
arXiv2024

Solar cycle variability induced by stochastic fluctuations of BMR properties and at different amounts of dynamo supercriticality

Understanding the irregular variation of the solar cycle is crucial due to its significant impact on global climates and the heliosphere. Since the polar magnetic field determines the amplitude of the next solar cycle, variations in the polar field can lead to fluctuations in the solar cycle. We have explored the variability of the solar cycle at different levels of dynamo supercriticality. We observe that the variability depends on the dynamo operation regime, with the near-critical regime exhibiting more variability than the supercritical regime. Furthermore, we have explored the effects of the irregular BMR properties (emergence rate, latitude, tilt, and flux) on the polar field and the solar cycle. We find that they all produce considerable variation in the solar cycle; however, the variation due to the tilt scatter is the largest.

astro-ph.SR↗
arXiv2023

Alpha Elimination: Using Deep Reinforcement Learning to Reduce Fill-In during Sparse Matrix Decomposition

A large number of computational and scientific methods commonly require decomposing a sparse matrix into triangular factors as LU decomposition. A common problem faced during this decomposition is that even though the given matrix may be very sparse, the decomposition may lead to a denser triangular factors due to fill-in. A significant fill-in may lead to prohibitively larger computational costs and memory requirement during decomposition as well as during the solve phase. To this end, several heuristic sparse matrix reordering methods have been proposed to reduce fill-in before the decomposition. However, finding an optimal reordering algorithm that leads to minimal fill-in during such decomposition is known to be a NP-hard problem. A reinforcement learning based approach is proposed for this problem. The sparse matrix reordering problem is formulated as a single player game. More specifically, Monte-Carlo tree search in combination with neural network is used as a decision making algorithm to search for the best move in our game. The proposed method, alphaElimination is found to produce significantly lesser non-zeros in the LU decomposition as compared to existing state-of-the-art heuristic algorithms with little to no increase in overall running time of the algorithm. The code for the project will be publicly available here\footnote{\url{https://github.com/misterpawan/alphaEliminationPaper}}.

cs.LG↗
arXiv2023

Enhancing ML model accuracy for Digital VLSI circuits using diffusion models: A study on synthetic data generation

Generative AI has seen remarkable growth over the past few years, with diffusion models being state-of-the-art for image generation. This study investigates the use of diffusion models in generating artificial data generation for electronic circuits for enhancing the accuracy of subsequent machine learning models in tasks such as performance assessment, design, and testing when training data is usually known to be very limited. We utilize simulations in the HSPICE design environment with 22nm CMOS technology nodes to obtain representative real training data for our proposed diffusion model. Our results demonstrate the close resemblance of synthetic data using diffusion model to real data. We validate the quality of generated data, and demonstrate that data augmentation certainly effective in predictive analysis of VLSI design for digital circuits.

cs.LG↗
arXiv2023

Exploring the reliability of polar field rise rate as a precursor for an early prediction of solar cycle

The prediction of the strength of an upcoming solar cycle has been a long-standing challenge in the field of solar physics. The inherent stochastic nature of the underlying solar dynamo makes the strength of the solar cycle vary in a wide range. Till now, the polar precursor methods and the dynamo simulations, that use the strength of the polar field at the cycle minimum to predict the strength of the following cycle has gained reasonable consensus by providing convergence in the predictions for solar cycles 24 and 25. Recently, it has been shown that just by using the observed correlation of the polar field rise rate with the peak of the polar field at the cycle minimum and the amplitude of the following cycle, a reliable prediction can be made much earlier than the cycle minimum. In this work, we perform surface flux transport (SFT) simulations to explore the robustness of this correlation against the stochastic fluctuations of BMR tilt properties including anti-Joy and anti-Hale type anomalous BMRs, and against the variation of meridional flow speed. We find that the observed correlation is a robust feature of the solar cycles and thus it can be utilized for a reliable prediction of solar cycle much earlier than the cycle minimum, the usual landmark of the solar cycle prediction.

astro-ph.SR↗
arXiv2023

Constraining the FRB mechanism from scintillation in the host galaxy

Most FRB models can be divided into two groups based on the distance of the radio emission region from the central engine. The first group of models, the so-called `nearby' or magnetospheric models, invoke FRB emission at distances of 10$^9$ cm or less from the central engine, while the second `far-away' models involve emission from distances of 10$^{11}$ cm or greater. The lateral size for the emission region for the former class of models ($\lesssim$ 10$^7$ cm) is much smaller than the second class of models ($\gtrsim 10^9$ cm). We propose that an interstellar scattering screen in the host galaxy is well-suited to differentiate between the two classes of models, particularly based on the level of modulations in the observed intensity with frequency, in the regime of strong diffractive scintillation. This is because the diffractive length scale for the host galaxy's ISM scattering screen is expected to lie between the transverse emission-region sizes for the `nearby' and the `far-away' class of models. Determining the strength of flux modulation caused by scintillation (scintillation modulation index) across the scintillation bandwidth ($\sim 1/2πδt_s$) would provide a strong constraint on the FRB radiation mechanism when the scatter broadening ($δt_s$) is shown to be from the FRB host galaxy. The scaling of the scintillation bandwidth as $\sim ν^{4.4}$ may make it easier to determine the modulation index at $\gtrsim$ 1 GHz.

astro-ph.HE↗
arXiv2023

Riemannian Hamiltonian methods for min-max optimization on manifolds

In this paper, we study min-max optimization problems on Riemannian manifolds. We introduce a Riemannian Hamiltonian function, minimization of which serves as a proxy for solving the original min-max problems. Under the Riemannian Polyak--Łojasiewicz condition on the Hamiltonian function, its minimizer corresponds to the desired min-max saddle point. We also provide cases where this condition is satisfied. For geodesic-bilinear optimization in particular, solving the proxy problem leads to the correct search direction towards global optimality, which becomes challenging with the min-max formulation. To minimize the Hamiltonian function, we propose Riemannian Hamiltonian methods (RHM) and present their convergence analyses. We extend RHM to include consensus regularization and to the stochastic setting. We illustrate the efficacy of the proposed RHM in applications such as subspace robust Wasserstein distance, robust training of neural networks, and generative adversarial networks.

math.OC↗
arXiv2023

Predicting Contact Angle Hysteresis on Surfaces with Randomly and Periodically Distributed Cylindrical Pillars via Energy Dissipation

Hypothesis: Understanding contact angle hysteresis on rough surfaces is important as most industrially relevant and naturally occurring surfaces possess some form of random or structured roughness. We hypothesise that hysteresis originates from the energy dissipation during the $\textit{stick-slip}$ motion of the contact line and that this energy dissipation is key to developing a predictive equation for hysteresis. Experiments: We measured hysteresis on surfaces with randomly distributed and periodically arranged microscopic cylindrical pillars for a variety of different liquids in air. The inherent (flat surface) contact angles tested range from lyophilic ($θ_{\rm{e}}=33.8^{\circ}$) to lyophobic ($θ_{\rm{e}} = 112.0^{\circ}$). Findings: A new methodology for calculating the average advancing and receding contact angles on random surfaces is presented. Also, the correlations for roughness-induced energy dissipation were derived, and a predictive equation for the advancing and receding contact angles during homogeneous (Wenzel) wetting on random surfaces is presented. Significantly, equations that predict the onset of the alternate wetting conditions of hemiwicking, split-advancing, split-receding and heterogeneous (Cassie) wetting are also derived, thus defining the range of validity for the derived homogeneous wetting equation. A novel feature 'cluster' concept is introduced which explains the measurably higher hysteresis exhibited by structured surfaces compared to random surfaces observed experimentally.

physics.flu-dyn↗
arXiv2023

marl-jax: Multi-Agent Reinforcement Leaning Framework

Recent advances in Reinforcement Learning (RL) have led to many exciting applications. These advancements have been driven by improvements in both algorithms and engineering, which have resulted in faster training of RL agents. We present marl-jax, a multi-agent reinforcement learning software package for training and evaluating social generalization of the agents. The package is designed for training a population of agents in multi-agent environments and evaluating their ability to generalize to diverse background agents. It is built on top of DeepMind's JAX ecosystem~\cite{deepmind2020jax} and leverages the RL ecosystem developed by DeepMind. Our framework marl-jax is capable of working in cooperative and competitive, simultaneous-acting environments with multiple agents. The package offers an intuitive and user-friendly command-line interface for training a population and evaluating its generalization capabilities. In conclusion, marl-jax provides a valuable resource for researchers interested in exploring social generalization in the context of MARL. The open-source code for marl-jax is available at: \href{https://github.com/kinalmehta/marl-jax}{https://github.com/kinalmehta/marl-jax}

cs.MA↗
arXiv2023

Probing the variations in the timing of the Sun's polar magnetic field reversals through observations and surface flux transport simulations

The polar field reversal is a crucial process in the cyclic evolution of the large-scale magnetic field of the Sun.Various important characteristics of a solar cycle, such as its duration and strength, and also the cycle predictability, are determined by the polar field reversal time. While the regular measurements of solar magnetic field have been accumulated for more than half a century, there is no consensus in the heliophysics community concerning the interpretation of the Sun's polar field measurements and especially the determination of polar field reversal time. There exists a severe problem of non-reproducibility in the reported results even from studies of the same observational dataset, and this causes an obstacle to make more accurate forecasts of solar cycle. Here, we analyze the solar magnetograms from four instruments for the last four cycles, to provide a more correct interpretation of the polar field observations and to find more accurate time of the reversals. We show the absence of triple (multipolar) reversals in Cycles 21 - 24, significant variations in the time interval between reversals in the hemispheres and in the time interval between a reversal and a cycle beginning. In order to understand the origin of the reversal time variation, we perform Surface Flux Transport (SFT) simulations and find out that the presence of the 'anomalous' bipolar magnetic regions (BMRs) in different phases of a cycle can cause cycle-to-cycle variations of the reversal time within the similar range found in observations.

astro-ph.SR↗
arXiv2023

Ultra-fast Vacancy Migration: A Novel Approach for Synthesizing Sub-10 nm Crystalline Transition Metal Dichalcogenide Nanocrystals

Two-dimensional materials, such as transition metal dichalcogenides (TMDCs), have the potential to revolutionize the field of electronics and photonics due to their unique physical and structural properties. This research presents a novel method for synthesizing crystalline TMDCs crystals with < 10 nm size using ultra-fast migration of vacancies at elevated temperatures. Through in-situ and ex-situ processing and using atomic-level characterization techniques, we analyze the shape, size, crystallinity, composition, and strain distribution of these nanocrystals. These nanocrystals exhibit electronic structure signatures that differ from the 2D bulk i.e., uniform mono and multilayers. Further, our in-situ, vacuum-based synthesis technique allows observation and comparison of defect and phase evolution in these crystals formed under van der Waals heterostructure confinement versus unconfined conditions. Overall, this research demonstrates a solid-state route to synthesizing uniform nanocrystals of TMDCs and lays the foundation for materials science in confined 2D spaces under extreme conditions.

cond-mat.mtrl-sci↗
arXiv2023

The LIGO HET Response (LIGHETR) Project to Discover and Spectroscopically Follow Optical Transients Associated with Neutron Star Mergers

The LIGO HET Response (LIGHETR) project is an enterprise to follow up optical transients (OT) discovered as gravitational wave merger sources by the LIGO/Virgo collaboration (LVC). Early spectroscopy has the potential to constrain crucial parameters such as the aspect angle. The LIGHETR collaboration also includes the capacity to model the spectroscopic evolution of mergers to facilitate a real-time direct comparison of models with our data. The principal facility is the Hobby-Eberly Telescope. LIGHETR uses the massively-replicated VIRUS array of spectrographs to search for associated OTs and obtain early blue spectra and in a complementary role, the low-resolution LRS-2 spectrograph is used to obtain spectra of viable candidates as well as a densely-sampled series of spectra of true counterparts. Once an OT is identified, the anticipated cadence of spectra would match or considerably exceed anything achieved for GW170817 = AT2017gfo for which there were no spectra in the first 12 hours and thereafter only roughly once daily. We describe special HET-specific software written to facilitate the program and attempts to determine the flux limits to undetected sources. We also describe our campaign to follow up OT candidates during the third observational campaign of the LIGO and Virgo Scientific Collaborations. We obtained VIRUS spectroscopy of candidate galaxy hosts for 5 LVC gravitational wave events and LRS-2 spectra of one candidate for the OT associated with S190901ap. We identified that candidate, ZTF19abvionh = AT2019pip, as a possible Wolf-Rayet star in an otherwise unrecognized nearby dwarf galaxy.

astro-ph.HE↗
arXiv2023

Structured Low-Rank Tensor Learning

We consider the problem of learning low-rank tensors from partial observations with structural constraints, and propose a novel factorization of such tensors, which leads to a simpler optimization problem. The resulting problem is an optimization problem on manifolds. We develop first-order and second-order Riemannian optimization algorithms to solve it. The duality gap for the resulting problem is derived, and we experimentally verify the correctness of the proposed algorithm. We demonstrate the algorithm on nonnegative constraints and Hankel constraints.

cs.LG↗
arXiv2023

Nonnegative Low-Rank Tensor Completion via Dual Formulation with Applications to Image and Video Completion

Recent approaches to the tensor completion problem have often overlooked the nonnegative structure of the data. We consider the problem of learning a nonnegative low-rank tensor, and using duality theory, we propose a novel factorization of such tensors. The factorization decouples the nonnegative constraints from the low-rank constraints. The resulting problem is an optimization problem on manifolds, and we propose a variant of Riemannian conjugate gradients to solve it. We test the proposed algorithm across various tasks such as colour image inpainting, video completion, and hyperspectral image completion. Experimental results show that the proposed method outperforms many state-of-the-art tensor completion algorithms.

cs.CV↗
arXiv2023

Effects of Spectral Normalization in Multi-agent Reinforcement Learning

A reliable critic is central to on-policy actor-critic learning. But it becomes challenging to learn a reliable critic in a multi-agent sparse reward scenario due to two factors: 1) The joint action space grows exponentially with the number of agents 2) This, combined with the reward sparseness and environment noise, leads to large sample requirements for accurate learning. We show that regularising the critic with spectral normalization (SN) enables it to learn more robustly, even in multi-agent on-policy sparse reward scenarios. Our experiments show that the regularised critic is quickly able to learn from the sparse rewarding experience in the complex SMAC and RWARE domains. These findings highlight the importance of regularisation in the critic for stable learning.

cs.LG↗