arXiv subjects

Manoj Kumar

Publications and source records attributed to Manoj Kumar.

At least 55 records · Page 3Linked to original sources

arXiv2022

Minimizing the phase structure of quark mass matrices

Fritzsch-Xing matrices are a particular class of texture 4 zero hermitian quark mass matrices, known to be successful in accommodating the quark mixing data. In the present work, it is shown that these texture 4-zero matrices with only one phase parameter, unlike the usually considered two phase parameters, are not only consistent with the latest experimental quark mixing data, but also predict the CP violation parameters, $J$ and corresponding phase $δ$, in agreement with the recent global analyses. We also show that the mass matrix elements do not exhibit a strong hierarchy and there is a strong correlation between some of the mass matrix elements of up and down sector. A precision measurement of $δ$ as well as small quark masses would have the potential to constrain the phase structure of the matrices further.

hep-ph↗

arXiv2022

Titchmarsh theorems on Damek-Ricci spaces via moduli of continuity of higher order

A classical theorem of Titchmarsh relates the $L^2$-Lipschitz functions and decay of the Fourier transform of the functions. In this note, we prove the Titchmarsh theorem for Damek-Ricci space (also known as harmonic $NA$ groups) via moduli of continuity of higher orders. We also prove an analogue of another Titchmarsh theorem which provides integrability properties of the Fourier transform for functions in the Hölder Lipschitz spaces.

math.FA↗

arXiv2022

Quasi-exact ground-state algorithm for the random-field Potts model

The use of combinatorial optimization algorithms has contributed substantially to the major progress that has occurred in recent years in the understanding of the physics of disordered systems, such as the random-field Ising model. While for this system exact ground states can be computed efficiently in polynomial time, the related random-field Potts model is {\em NP\} hard computationally. While thus exact ground states cannot be computed for large systems in this case, approximation schemes based on graph cuts and related techniques can be used. Here we show how a combination of such methods with repeated runs allows for a systematic extrapolation of relevant system properties to the ground state. The method is benchmarked on a special class of disorder samples for which exact ground states are available.

cond-mat.dis-nn↗

arXiv2022

IISERB Brains at SemEval 2022 Task 6: A Deep-learning Framework to Identify Intended Sarcasm in English

This paper describes the system architectures and the models submitted by our team "IISERBBrains" to SemEval 2022 Task 6 competition. We contested for all three sub-tasks floated for the English dataset. On the leader-board, wegot19th rank out of43 teams for sub-taskA, the 8th rank out of22 teams for sub-task B,and13th rank out of 16 teams for sub-taskC. Apart from the submitted results and models, we also report the other models and results that we obtained through our experiments after organizers published the gold labels of their evaluation data

cs.CL↗

arXiv2021

Skillful Twelve Hour Precipitation Forecasts using Large Context Neural Networks

The problem of forecasting weather has been scientifically studied for centuries due to its high impact on human lives, transportation, food production and energy management, among others. Current operational forecasting models are based on physics and use supercomputers to simulate the atmosphere to make forecasts hours and days in advance. Better physics-based forecasts require improvements in the models themselves, which can be a substantial scientific challenge, as well as improvements in the underlying resolution, which can be computationally prohibitive. An emerging class of weather models based on neural networks represents a paradigm shift in weather forecasting: the models learn the required transformations from data instead of relying on hand-coded physics and are computationally efficient. For neural models, however, each additional hour of lead time poses a substantial challenge as it requires capturing ever larger spatial contexts and increases the uncertainty of the prediction. In this work, we present a neural network that is capable of large-scale precipitation forecasting up to twelve hours ahead and, starting from the same atmospheric state, the model achieves greater skill than the state-of-the-art physics-based models HRRR and HREF that currently operate in the Continental United States. Interpretability analyses reinforce the observation that the model learns to emulate advanced physics principles. These results represent a substantial step towards establishing a new paradigm of efficient forecasting with neural networks.

cs.LG↗

arXiv2021

Encrypted Data Processing

In this paper, we present a comprehensive architecture for confidential computing, which we show to be general purpose and quite efficient. It executes the application as is, without any added burden or discipline requirements from the application developers. Furthermore, it does not require the trust of system software at the computing server and does not impose any added burden on the communication subsystem. The proposed Encrypted Data Processing (EDAP) architecture accomplishes confidentiality, authenticity, and freshness of the key-based cryptographic data protection by adopting data encryption with a multi-level key protection scheme. It guarantees that the user data is visible only in non-privileged mode to a designated program trusted by the data owner on a designated hardware, thus protecting the data from an untrusted hardware, hypervisor, OS, or other users' applications. The cryptographic keys and protocols used for achieving these confidential computing requirements are described in a use case example. Encrypting and decrypting data in an EDAP-enabled processor can lead to performance degradation as it adds cycle time to the overall execution. However, our simulation result shows that the slowdown is only 6% on average across a collection of commercial workloads when the data encryption engine is placed between the L1 and L2 cache. We demonstrate that the EDAP architecture is valuable and practicable in the modern cloud environment for confidential computing. EDAP delivers a zero trust model of computing where the user software does not trust system software and vice versa.

cs.CR↗

arXiv2021

Domain Growth and Aging in the Random Field XY Model: A Monte Carlo Study

We use large-scale Monte Carlo simulations to obtain comprehensive results for domain growth and aging in the random field XY model in dimensions $d=2,3$. After a deep quench from the paramagnetic phase, the system orders locally via annihilation of topological defects, i.e., vortices and anti-vortices. The evolution morphology of the system is characterized by the correlation function and the structure factor of the magnetization field. We find that these quantities obey dynamical scaling, and their scaling function is independent of the disorder strength $Δ$. However, the scaling form of the autocorrelation function is found to be dependent on $Δ$, i.e., superuniversality is violated. The large-$t$ behavior of the autocorrelation function is explored by studying aging and autocorrelation exponents. We also investigate the characteristic growth law $L(t,Δ)$ in $d=2,3$, which shows an asymptotic logarithmic behavior: $L(t,Δ) \sim Δ^{-φ} (\ln t)^{1/ψ}$, with exponents $φ, ψ> 0$.

cond-mat.stat-mech↗

arXiv2021

Superconducting Properties of La$_{1.85}$Sr$_{0.15}$CuO$_4$-Sr$_2$IrO$_4$ multilayers

We study superconducting properties in multilayer thin films consisting of superconducting La$_{1.85}$Sr$_{0.15}$CuO$_4$ (LSCO) and Mott insulator Sr$_2$IrO$_4$ (SIO) and report enhanced superconductivity in optimized sample. These multilayer heterostructures show an increase in superconducting transition temperature ($T_C$) as compared to the single layer LSCO films. The temperature dependence of SIO single layer is also investigated under thermal activation, Arrhenius-type behaviour, and variable-range hopping mechanisms for different temperature regimes. The decrease in $T_C$ beyond an optimum thickness of LSCO in these multilayers is analyzed in the framework of a model based on the assumption of induced superconductivity in SIO-LSCO interface due to the doping of La and/or oxygen deficiencies into SIO layers

cond-mat.supr-con↗

arXiv2021

KEA: Tuning an Exabyte-Scale Data Infrastructure

Microsoft's internal big-data infrastructure is one of the largest in the world -- with over 300k machines running billions of tasks from over 0.6M daily jobs. Operating this infrastructure is a costly and complex endeavor, and efficiency is paramount. In fact, for over 15 years, a dedicated engineering team has tuned almost every aspect of this infrastructure, achieving state-of-the-art efficiency (>60% average CPU utilization across all clusters). Despite rich telemetry and strong expertise, faced with evolving hardware/software/workloads this manual tuning approach had reached its limit -- we had plateaued. In this paper, we present KEA, a multi-year effort to automate our tuning processes to be fully data/model-driven. KEA leverages a mix of domain knowledge and principled data science to capture the essence of our cluster dynamic behavior in a set of machine learning (ML) models based on collected system data. These models power automated optimization procedures for parameter tuning, and inform our leadership in critical decisions around engineering and capacity management (such as hardware and data center design, software investments, etc.). We combine "observational" tuning (i.e., using models to predict system behavior without direct experimentation) with judicious use of "flighting" (i.e., conservative testing in production). This allows us to support a broad range of applications that we discuss in this paper. KEA continuously tunes our cluster configurations and is on track to save Microsoft tens of millions of dollars per year. At the best of our knowledge, this paper is the first to discuss research challenges and practical learnings that emerge when tuning an exabyte-scale data infrastructure.

cs.DB↗

arXiv2021

Necessary and Sufficient Condition for Satisfiability of a Boolean Formula in CNF and its Implications on P versus NP problem

Boolean satisfiability problem has applications in various fields. An efficient algorithm to solve satisfiability problem can be used to solve many other problems efficiently. The input of satisfiability problem is a finite set of clauses. In this paper, properties of clauses have been studied. A type of clauses have been defined, called fully populated clauses, which contains each variable exactly once. A relationship between two unequal fully populated clauses has been defined, called sibling clauses. It has been found that, if one fully populated clause is false, for a truth assignment, then all it's sibling clauses will be true for the same truth assignment. Which leads to the necessary and sufficient condition for satisfiability of a boolean formula, in CNF. The necessary and sufficient condition has been used to develop a novel algorithm to solve boolean satisfiability problem in polynomial time, which implies, P equals NP. Further, some optimisations have been provided that can be integrated with the algorithm for better performance.

cs.CC↗

arXiv2021

Coexistence of coarsening and mean field relaxation in the long-range Ising chain

We study the kinetics after a low temperature quench of the one-dimensional Ising model with long range interactions between spins at distance $r$ decaying as $r^{-α}$. For $α=0$, i.e. mean field, all spins evolve coherently quickly driving the system towards a magnetised state. In the weak long range regime with $α>1$ there is a coarsening behaviour with competing domains of opposite sign without development of magnetisation. For strong long range, i.e. $0<α<1$, we show that the system shows both features, with probability $P_α(N)$ of having the latter one, with the different limiting behaviours $\lim _{N\to \infty}P_α(N)=0$ (at fixed $α<1$) and $\lim _{α\to 1}P_α(N)=1$ (at fixed finite $N$). We discuss how this behaviour is a manifestation of an underlying dynamical scaling symmetry due to the presence of a single characteristic time $τ_α(N)\sim N^α$.

cond-mat.stat-mech↗

arXiv2021

Optimal Birth Control for a Size-Structured Population Model with Diffusion

This work addresses the optimal birth control problem for invasive species in a spatial environment. We apply the method of semigroups to qualitatively analyze a size-structured population model in which individuals occupy a position in a spatial environment. With insect population in mind, we study the optimal control problem which takes fertility rate as a control variable. With the help of adjoint system, we derive optimality conditions. We obtain the optimality conditions by fixing the birth rate on three different sets. Using Ekeland's variational principle, the existence, and uniqueness of optimal birth controller to the given population model which minimizes a given cost functional is shown. Outcomes of our article are new and complement the existing ones.

math.OC↗

arXiv2021

Colorization Transformer

We present the Colorization Transformer, a novel approach for diverse high fidelity image colorization based on self-attention. Given a grayscale image, the colorization proceeds in three steps. We first use a conditional autoregressive transformer to produce a low resolution coarse coloring of the grayscale image. Our architecture adopts conditional transformer layers to effectively condition grayscale input. Two subsequent fully parallel networks upsample the coarse colored low resolution image into a finely colored high resolution image. Sampling from the Colorization Transformer produces diverse colorings whose fidelity outperforms the previous state-of-the-art on colorising ImageNet based on FID results and based on a human evaluation in a Mechanical Turk test. Remarkably, in more than 60% of cases human evaluators prefer the highest rated among three generated colorings over the ground truth. The code and pre-trained checkpoints for Colorization Transformer are publicly available at https://github.com/google-research/google-research/tree/master/coltran

cs.CV↗

arXiv2021

ProtoDA: Efficient Transfer Learning for Few-Shot Intent Classification

Practical sequence classification tasks in natural language processing often suffer from low training data availability for target classes. Recent works towards mitigating this problem have focused on transfer learning using embeddings pre-trained on often unrelated tasks, for instance, language modeling. We adopt an alternative approach by transfer learning on an ensemble of related tasks using prototypical networks under the meta-learning paradigm. Using intent classification as a case study, we demonstrate that increasing variability in training tasks can significantly improve classification performance. Further, we apply data augmentation in conjunction with meta-learning to reduce sampling bias. We make use of a conditional generator for data augmentation that is trained directly using the meta-learning objective and simultaneously with prototypical networks, hence ensuring that data augmentation is customized to the task. We explore augmentation in the sentence embedding space as well as prototypical embedding space. Combining meta-learning with augmentation provides upto 6.49% and 8.53% relative F1-score improvements over the best performing systems in the 5-shot and 10-shot learning, respectively.

cs.CL↗

arXiv2020

Analysis of steady state solutions to an age structured SEQIR model with optimal vaccination

Quarantine of those individuals who are suspected of being infected is one of the intervention measures to contain the spread of an infectious disease. We propose an age structured SEQIR (S-Susceptible, E-Exposed, Q-Quarantine, I-Infected, R-Recovered) model with vaccination of susceptible and exposed individuals. Firstly without vaccination, basic reproduction number is derived by using the appearance of endemic steady state. With an appropriate Liapunov function, stability of disease free equilibrium point is checked. Vaccinating a population with reduction in number of infected individuals at minimal cost is considered as an optimization problem. We show that the vaccination strategy is concentrated on atmost three age classes.

math.AP↗

arXiv2020

Multi-Scale Speaker Diarization With Neural Affinity Score Fusion

Identifying the identity of the speaker of short segments in human dialogue has been considered one of the most challenging problems in speech signal processing. Speaker representations of short speech segments tend to be unreliable, resulting in poor fidelity of speaker representations in tasks requiring speaker recognition. In this paper, we propose an unconventional method that tackles the trade-off between temporal resolution and the quality of the speaker representations. To find a set of weights that balance the scores from multiple temporal scales of segments, a neural affinity score fusion model is presented. Using the CALLHOME dataset, we show that our proposed multi-scale segmentation and integration approach can achieve a state-of-the-art diarization performance.

eess.AS↗

arXiv2020

Transport, correlations, and chaos in a classical disordered anharmonic chain

We explore transport properties in a disordered nonlinear chain of classical harmonic oscillators and thereby identify a regime exhibiting behavior analogous to that seen in quantum many-body-localized systems. Through extensive numerical simulations of this system connected at its ends to heat baths at different temperatures, we computed the heat current and the temperature profile in the nonequilibrium steady state as a function of system size $N$, disorder strength $Δ$, and temperature $T$. The conductivity $κ_N$, obtained for finite length ($N$) systems, saturates to a value $κ_\infty >0$ in the large $N$ limit, for all values of disorder strength $Δ$ and temperature $T>0$. We show evidence that for any $Δ>0$ the conductivity goes to zero faster than any power of $T$ in the $(T/Δ) \to 0$ limit, and find that the form $κ_\infty \sim e^{-B |\ln(C Δ/T)|^3}$ fits our data. This form has earlier been suggested by a theory based on the dynamics of multi-oscillator chaotic islands. The finite-size effect can be $κ_N < κ_{\infty}$ due to boundary resistance when the bulk conductivity is high (the weak disorder case), or $κ_N > κ_{\infty}$ due to direct bath-to-bath coupling through bulk localized modes when the bulk is weakly conducting (the strong disorder case). We also present results on equilibrium dynamical correlation functions and on the role of chaos on transport properties. Finally, we explore the differences in the growth and propagation of chaos in the weak and strong chaos regimes by studying the classical version of the Out-of-Time-Ordered-Commutator.

cond-mat.stat-mech↗

arXiv2020

Designing Neural Speaker Embeddings with Meta Learning

Neural speaker embeddings trained using classification objectives have demonstrated state-of-the-art performance in multiple applications. Typically, such embeddings are trained on an out-of-domain corpus on a single task e.g., speaker classification, albeit with a large number of classes (speakers). In this work, we reformulate embedding training under the meta-learning paradigm. We redistribute the training corpus as an ensemble of multiple related speaker classification tasks, and learn a representation that generalizes better to unseen speakers. First, we develop an open source toolkit to train x-vectors that is matched in performance with pre-trained Kaldi models for speaker diarization and speaker verification applications. We find that different bottleneck layers in the architecture variedly favor different applications. Next, we use two meta-learning strategies, namely prototypical networks and relation networks, to improve over the x-vector embeddings. Our best performing model achieves a relative improvement of 12.37% and 7.11% in speaker error on the DIHARD II development corpus and the AMI meeting corpus, respectively. We analyze improvements across different domains in the DIHARD corpus. Notably, on the challenging child speech domain, we study the relation between child age and the diarization performance. Further, we show reductions in equal error rate for speaker verification on the SITW corpus (7.68%) and the VOiCES challenge corpus (8.78%). We observe that meta-learning particularly offers benefits in challenging acoustic conditions and recording setups encountered in these corpora. Our experiments illustrate the applicability of meta-learning as a generalized learning paradigm for training deep neural speaker embeddings.

eess.AS↗