Searcharxiv⌕ Search

arXiv subjects

Sourav Dutta

Publications and source records attributed to Sourav Dutta.

At least 37 records · Page 2Linked to original sources
arXiv2024

RSVP: Beyond Weisfeiler Lehman Graph Isomorphism Test

Graph isomorphism, a classical algorithmic problem, determines whether two input graphs are structurally identical or not. Interestingly, it is one of the few problems that is not yet known to belong to either the P or NP-complete complexity classes. As such, intelligent search-space pruning based strategies were proposed for developing isomorphism testing solvers like nauty and bliss, which are still, unfortunately, exponential in the worst-case scenario. Thus, the polynomial-time Weisfeiler-Lehman (WL) isomorphism testing heuristic, based on colour refinement, has been widely adopted in the literature. However, WL fails for multiple classes of non-isomorphic graph instances such as strongly regular graphs, block structures, and switched edges, among others. In this paper, we propose a novel polynomial-time graph isomorphism testing heuristic, RSVP, and depict its enhanced discriminative power compared to the Weisfeiler-Lehman approach for several challenging classes of graphs. Bounded by a run-time complexity of O(m^2+mn^2+n^3) (where n and m are the number of vertices and edges respectively), we show that RSVP can identify non-isomorphism in several 'hard' graph instance classes including Miyazaki, Paulus, cubic hypohamiltonian, strongly regular, Latin series and Steiner triple system graphs, where the 3-WL test fails. Similar to the WL test, our proposed algorithm is prone to only one-sided errors, where isomorphic graphs will never be determined to be non-isomorphic, although the reverse can happen.

cs.DS↗
arXiv2024

A description of classical and quantum cosmology for a single scalar field torsion gravity

In the background of homogeneous and isotropic flat FLRW space-time, both classical and quantum cosmology has been studied for teleparallel dark energy (DE) model. Using Noether symmetry analysis, not only the symmetry vector but also the coupling function in the Lagrangian and the potential of the scalar field has been determined. Also symmetry analysis identifies a cyclic variable in the Lagrangian along the symmetry vector and as a result the Lagrangian simplifies to a great extend so that classical solution is obtained. Subsequently, in quantum cosmology Wheeler-DeWitt(WD) equation has been constructed and the quantum version of the conserved momenta corresponding to Noether symmetry identifies the periodic part of the wave function of the universe and as a result the Wheeler-DeWitt equation becomes solvable. Finally, quantum description shows finite non-zero probability at the classical big-bang singularity.

gr-qc↗
arXiv2024

Classical and Quantum Cosmology in Einstein-aether Scalar-tensor gravity: Noether Symmetry Analysis

The present work deals with Einstein-aether Scalar tensor gravity in the background of homogeneous and isotropic flat FLRW space-time model. The Noether symmetry vector identifies a transformation in the augmented space so that the field equations become solvable. The cosmological solutions are analyzed from the observational point of view. Finally, for quantum cosmology, the Wheeler-DeWitt (WD) has been formulated and solutions have been determined by identifying the periodic nature of the wave function using conserved (Noether) charge.

gr-qc↗
arXiv2024

Quantum Cosmology in Coupled Brans-Dicke Gravity: A Noether Symmetry Analysis

The present work deals with a multi-field cosmological model in a spatially flat FLRW space-time geometry. The usual Brans-Dicke(BD) field and another scalar field are minimally coupled to gravity while they interact with each other through the Kinetic terms. {The main aim of the present work is to examine whether the model is compatible with cosmic observations. So cosmological solutions are obtained using symmetry analysis only.} By imposing Noether Symmetry to the Lagrangian of the system the potential of the scalar field as well as the coupling function has been determined. The classical solutions are determined after simplifying the Lagrangian using cyclic variables. Finally, Wheeler-DeWitt(WD) equation in quantum cosmology has been formulated and conserved momenta corresponding to Noether symmetry shows the periodic part of the wave function and it helps to have the complete integral for the wave function.

gr-qc↗
arXiv2024

Improved Out-of-Scope Intent Classification with Dual Encoding and Threshold-based Re-Classification

Detecting out-of-scope user utterances is essential for task-oriented dialogues and intent classification. Current methodologies face difficulties with the unpredictable distribution of outliers and often rely on assumptions about data distributions. We present the Dual Encoder for Threshold-Based Re-Classification (DETER) to address these challenges. This end-to-end framework efficiently detects out-of-scope intents without requiring assumptions on data distributions or additional post-processing steps. The core of DETER utilizes dual text encoders, the Universal Sentence Encoder (USE) and the Transformer-based Denoising AutoEncoder (TSDAE), to generate user utterance embeddings, which are classified through a branched neural architecture. Further, DETER generates synthetic outliers using self-supervision and incorporates out-of-scope phrases from open-domain datasets. This approach ensures a comprehensive training set for out-of-scope detection. Additionally, a threshold-based re-classification mechanism refines the model's initial predictions. Evaluations on the CLINC-150, Stackoverflow, and Banking77 datasets demonstrate DETER's efficacy. Our model outperforms previous benchmarks, increasing up to 13% and 5% in F1 score for known and unknown intents on CLINC-150 and Stackoverflow, and 16% for known and 24% % for unknown intents on Banking77. The source code has been released at https://github.com/Hossam-Mohammed-tech/Intent_Classification_OOS.

cs.CL↗
arXiv2024

Late time dynamics in SUSY saddle-dominated scrambling through higher-point OTOC

In this article, we study the scrambling dynamics in supersymmetric quantum mechanical systems. The eigenstate representation of such supersymmetric systems allows us to present an explicit form of the $2N$-point out-of-time-order correlator (OTOC) using two equivalent formalisms viz. "Tensor Product formalism" and "Partner Hamiltonian formalism". We analytically compute the $2N$-point OTOC for the supersymmetric 1D harmonic oscillator and find that the result is in exact agreement with that of the OTOC of the 1D bosonic harmonic oscillator system. The higher-point OTOC is a more sensitive measure of scrambling than the usual 4-point OTOC. To demonstrate this feature, we consider a supersymmetric sextic 1D oscillator for which the bosonic partner system has an unstable saddle in the phase space, which is absent in the fermionic counterpart. For such a system we show that the bosonic, the fermionic as well as the supersymmetric OTOC exhibit similar dynamics due to supersymmetry constraints. Finally, we illustrate the late-time oscillatory behaviour of higher-point OTOC for saddle-dominated scrambling and anharmonic oscillator systems and propose it to be a probe of late-time dynamics in non-chaotic systems that exhibit fast early-time scrambling.

quant-ph↗
arXiv2024

A versatile apparatus for simultaneous trapping of multiple species of ultracold atoms and ions to enable studies of low energy collisions and cold chemistry

We describe an apparatus where many species of ultracold atoms can be simultaneously trapped and overlapped with many species of ions in a Paul trap. Several design innovations are made to increase the versatility of the apparatus while keeping the size and cost reasonable. We demonstrate the operation of a 3-dimensional (3D) magneto-optical trap (MOT) of $^7$Li using a single external cavity diode laser. The $^7$Li MOT is loaded from an atomic beam, with atoms slowed using a Zeeman slower designed to work simultaneously for Li and Sr. The operation of a 3D MOT of $^{133}$Cs, loaded from a 2D MOT, is demonstrated and provisions for MOTs of Rb and K in the same vacuum manifold exist. We demonstrate the trapping of $^7$Li$^+$ and $^{133}$Cs$^+$ at different settings of the Paul trap and their detection using an integrated time-of-flight mass spectrometer. We present results on low energy neutral-neutral collisions ($^{133}$Cs-$^{133}$Cs, $^7$Li-$^7$Li and $^{133}$Cs-$^7$Li collisions) and charge-neutral collisions ($^{133}$Cs$^+$-$^{133}$Cs and $^7$Li$^+$-$^7$Li collisions). We show evidence of sympathetic cooling of $^7$Li$^+$ ($^{133}$Cs$^+$) due to collisions with the ultracold $^7$Li ($^{133}$Cs).

physics.atom-ph↗
arXiv2023

AdaSent: Efficient Domain-Adapted Sentence Embeddings for Few-Shot Classification

Recent work has found that few-shot sentence classification based on pre-trained Sentence Encoders (SEs) is efficient, robust, and effective. In this work, we investigate strategies for domain-specialization in the context of few-shot sentence classification with SEs. We first establish that unsupervised Domain-Adaptive Pre-Training (DAPT) of a base Pre-trained Language Model (PLM) (i.e., not an SE) substantially improves the accuracy of few-shot sentence classification by up to 8.4 points. However, applying DAPT on SEs, on the one hand, disrupts the effects of their (general-domain) Sentence Embedding Pre-Training (SEPT). On the other hand, applying general-domain SEPT on top of a domain-adapted base PLM (i.e., after DAPT) is effective but inefficient, since the computationally expensive SEPT needs to be executed on top of a DAPT-ed PLM of each domain. As a solution, we propose AdaSent, which decouples SEPT from DAPT by training a SEPT adapter on the base PLM. The adapter can be inserted into DAPT-ed PLMs from any domain. We demonstrate AdaSent's effectiveness in extensive experiments on 17 different few-shot sentence classification datasets. AdaSent matches or surpasses the performance of full SEPT on DAPT-ed PLM, while substantially reducing the training costs. The code for AdaSent is available.

cs.CL↗
arXiv2023

Observation of quantum interference of optical transition pathways in Doppler-free two-photon spectroscopy and implications for precision measurements

Doppler-free two-photon spectroscopy is a standard technique for precision measurement of transition frequencies of dipole-forbidden transitions. The accuracy of such measurements depends critically on fitting of the spectrum to an appropriate line shape model, often a Voigt profile which neglects the effect of quantum interference of optical transitions. Here, we report the observation of quantum interference of optical transition pathways in Doppler-free two-photon spectroscopy of the cesium 6S-7D transitions. The quantum interference manifests itself as asymmetric line shapes of the hyperfine lines of the 7D states, observed through spontaneous emission following excitation by a narrow-linewidth cw laser. The interference persists despite the lines being spectrally well-resolved. Ignoring the effect of quantum interference causes large systematic shifts in the determination of the line-centers, while accounting for it resolves the apparent line shift and enables the precise determination of hyperfine splitting in the 7D states. We calculate the spectral line shape including the effect of quantum interference and show that it agrees with the experimental observations. Our results have implications for precision measurements of hyperfine splittings, isotope shifts and transition frequencies, including those of the hydrogen S-S and S-D transitions.

physics.atom-ph↗
arXiv2023

Gradient Sparsification For Masked Fine-Tuning of Transformers

Fine-tuning pretrained self-supervised language models is widely adopted for transfer learning to downstream tasks. Fine-tuning can be achieved by freezing gradients of the pretrained network and only updating gradients of a newly added classification layer, or by performing gradient updates on all parameters. Gradual unfreezing makes a trade-off between the two by gradually unfreezing gradients of whole layers during training. This has been an effective strategy to trade-off between storage and training speed with generalization performance. However, it is not clear whether gradually unfreezing layers throughout training is optimal, compared to sparse variants of gradual unfreezing which may improve fine-tuning performance. In this paper, we propose to stochastically mask gradients to regularize pretrained language models for improving overall fine-tuned performance. We introduce GradDrop and variants thereof, a class of gradient sparsification methods that mask gradients during the backward pass, acting as gradient noise. GradDrop is sparse and stochastic unlike gradual freezing. Extensive experiments on the multilingual XGLUE benchmark with XLMR-Large show that GradDrop is competitive against methods that use additional translated data for intermediate pretraining and outperforms standard fine-tuning and gradual unfreezing. A post-analysis shows how GradDrop improves performance with languages it was not trained on, such as under-resourced languages.

cs.CL↗
arXiv2023

AI-assisted Improved Service Provisioning for Low-latency XR over 5G NR

Extended Reality (XR) is one of the most important 5G/6G media applications that will fundamentally transform human interactions. However, ensuring low latency, high data rate, and reliability to support XR services poses significant challenges. This letter presents a novel AI-assisted service provisioning scheme that leverages predicted frames for processing rather than relying solely on actual frames. This method virtually increases the network delay budget and consequently improves service provisioning, albeit at the expense of minor prediction errors. The proposed scheme is validated by extensive simulations demonstrating a multi-fold increase in supported XR users and also provides crucial network design insights.

cs.NI↗
arXiv2023

Attention over pre-trained Sentence Embeddings for Long Document Classification

Despite being the current de-facto models in most NLP tasks, transformers are often limited to short sequences due to their quadratic attention complexity on the number of tokens. Several attempts to address this issue were studied, either by reducing the cost of the self-attention computation or by modeling smaller sequences and combining them through a recurrence mechanism or using a new transformer model. In this paper, we suggest to take advantage of pre-trained sentence transformers to start from semantically meaningful embeddings of the individual sentences, and then combine them through a small attention layer that scales linearly with the document length. We report the results obtained by this simple architecture on three standard document classification datasets. When compared with the current state-of-the-art models using standard fine-tuning, the studied method obtains competitive results (even if there is no clear best model in this configuration). We also showcase that the studied architecture obtains better results when freezing the underlying transformers. A configuration that is useful when we need to avoid complete fine-tuning (e.g. when the same frozen transformer is shared by different applications). Finally, two additional experiments are provided to further evaluate the relevancy of the studied architecture over simpler baselines.

cs.CL↗
arXiv2023

Self-Distilled Quantization: Achieving High Compression Rates in Transformer-Based Language Models

We investigate the effects of post-training quantization and quantization-aware training on the generalization of Transformer language models. We present a new method called self-distilled quantization (SDQ) that minimizes accumulative quantization errors and outperforms baselines. We apply SDQ to multilingual models XLM-R-Base and InfoXLM-Base and demonstrate that both models can be reduced from 32-bit floating point weights to 8-bit integer weights while maintaining a high level of performance on the XGLUE benchmark. Our results also highlight the challenges of quantizing multilingual models, which must generalize to languages they were not fine-tuned on.

cs.CL↗
arXiv2023

Generalized Linear Models of T$_{90}$-T$_{50}$ relation to classify GRBs

Gamma-ray bursts (GRBs) can be classified with their linearly dependent parameters alongside the standard $T_{90}$ distribution. The Generalized linear mixture model(GLM) identifies the number of linear dependencies in a two-parameter space. Classically, GRBs are classified into two classes by the presence of bimodality in the histogram of T$_{90}$. However, additional classes and sub-classes of GRBs are fascinating topics to explore. In this work, we investigate the GRBs classes in the $ T_{90} {-}T_{50}$ plane using the Generalized Linear Models(GLM) for Fermi GBM and BATSE catalogs. This study shows five linear features for the Fermi GBM catalog and four linear features for the BATSE catalog, directing towards the possibility of more than two GRB classes.

astro-ph.HE↗
arXiv2023

Classical and quantum cosmology in $f(T)$-gravity theory: A Noether symmetry approach

In the framework of $f(T)$-gravity theory, classical and quantum cosmology has been studied in the present work for FLRW space-time model. The Noether symmetry, a point-like symmetry of the Lagrangian is used to the physical system and a specific functional form of $f(T)$ is determined. A point transformation in the 2D augmented space restricts one of the variable to be cyclic so that the Lagrangian as well as the field equations are simplified so that they are solvable. Lastly for quantum cosmology, the WD equation is constructed and possible solution has been evaluated.

gr-qc↗
arXiv2023

A stable 671 nm external cavity diode laser with output power exceeding 150 mW suitable for laser cooling of lithium atoms

We report the design and performance of a Littrow-type 671 nm External Cavity Diode Laser (ECDL) that delivers output power greater than 150 mW and features enhanced passive stability. The main body of the ECDL is constructed using titanium to minimize temperature related frequency drifts. The laser diode is mounted in a cylindrical mount that allows vertical adjustments while maintaining thermal contact with the temperature stabilized base plate. The wavelength tuning is achieved by horizontal displacement of the diffraction grating about an optimal pivot point. The compact design increases the robustness and passive stability of the ECDL and the stiff but light-weight diffraction grating-arm reduces the susceptibility to low-frequency mechanical vibrations. The linewidth of the ECDL is ~360 kHz. We use the 671 nm ECDL, without any additional power amplification, for laser cooling and trapping of lithium atoms in a magneto-optical trap. This simple, low-cost ECDL design using off-the-shelf laser diodes without anti-reflection coating can also be adapted to other wavelengths.

physics.optics↗
arXiv2023

Noether Symmetry analysis in Chameleon Field Cosmology

This work deals with chameleon field cosmology (a scalar field nonminimally coupled to cold dark matter) in the background of flat Friedmann-Lemaitre-Robertson-Walker (FLRW) space-time. Both classical and quantum cosmology have been investigated using Noether symmetry analysis of the underlying physical system. The Wheeler-DeWitt (WD) equation has been constructed on the minisuperspace and solutions have been obtained using conserved charge.

gr-qc↗
arXiv2023

Multiscalar field cosmological model and possible solutions using Noether symmetry approach

In this work, a cosmological model is considered having two scalar fields minimally coupled to gravity with a mixed kinetic term. The model is characterized by the coupling function and the potential function which are assumed to depend on one of the scalar fields. Instead of choosing these functions phenomenologically here, they are evaluated assuming the existence of Noether symmetry. By appropriate choice of a point transformation in the augmented space, one of the variables in the Lagrangian becomes cyclic and the evolution equations become much simpler to have solutions. Finally, the solutions are analyzed from cosmological view point.

gr-qc↗