arXiv subjects

Subhadeep Sarkar

Publications and source records attributed to Subhadeep Sarkar.

11 recordsLinked to original sources

arXiv2026

Revisiting the Electroweakino Sector of the Baryon Number Violating MSSM at the HL-LHC with Deep Neural Networks

We study the projected sensitivity of direct electroweakino production $pp \to \tilde\chi_1^{\pm} \tilde\chi_2^0$ at the HL-LHC in a simplified framework with wino-like, mass degenerate $\tilde\chi_1^{\pm}$ and $\tilde\chi_2^0$, and a bino-like lightest neutralino $\tilde\chi_1^0$, assuming R-parity violating~(RPV) through the baryon number violating $\lambda^{\prime \prime}_{112}u^c d^c d^c$ and $\lambda^{\prime \prime}_{113}u^c d^c b^c$ operators. We consider three channels with the $\lambda^{\prime \prime}_{112}u^c d^c d^c$ RPV operator: $Wh$ mediated $1\,\ell + 2\,b + \rm E{\!\!\!/}_T$, $Wh$ mediated $1\,\ell + (\geq 2\,j) + 2\, \gamma + \rm E{\!\!\!/}_T$, and $WZ$ mediated $3\ell + (\geq 2 j) + \rm E{\!\!\!/}_T$. In each channel, we train benchmark-specific multi-layer perceptrons (MLPs), analogous to signal-region classifiers, on the four-momenta of the final state particles along with a small set of higher-level observables to distinguish the signal from the dominant SM backgrounds. We find that the HL-LHC will be able to probe winos up to $\sim 900~$GeV, $\sim 780~$GeV, and $\sim 880~$GeV in the $Wh$ mediated $1\,\ell + 2\,b + \rm E{\!\!\!/}_T$, $Wh$ mediated $1\,\ell + (\geq 2\,j) + 2\, \gamma + \rm E{\!\!\!/}_T$, and $WZ$ mediated $3\ell + (\geq 2 j) + \rm E{\!\!\!/}_T$ channels, respectively, for $m_{\tilde\chi_1^0} \sim 50~$GeV, in the presence of $\lambda^{\prime \prime}_{112}u^c d^c d^c$ couplings, at $2\sigma$ sensitivity. In case the $\lambda^{\prime \prime}_{113}u^c d^c b^c$ operator is solely switched on, the projected sensitivity for winos reach up to $\sim 700~$GeV for $Wh$ mediated $1\,\ell + (\geq 1\,b)\, + (\geq 1j)\, + 2\, \gamma + \rm E{\!\!\!/}_T$ and $\sim 850~$GeV for the $WZ$ mediated $3\ell + (\geq 1 b) + \rm E{\!\!\!/}_T$ channel.

hep-ph↗

arXiv2025

Eliminating the Hidden Cost of Zone Management in ZNS SSDs

Zoned Namespace (ZNS) SSDs offer a new storage model that allows for high throughput and low-latency storage by eliminating device-side garbage collection. The ZNS interface exposes storage as append-only zones, thus enforcing host applications (e.g., database systems) to append, read, and garbage collect their pages. However, the storage abstraction of ZNS SSD hides the substantial differences across different ZNS SSD controller designs, which affects both the performance and predictability of host applications. We find that existing ZNS controllers exhibit (a) increased device-level write amplification (DLWA), (b) increased wear, and (c) increased interference with host I/O. We identify that (i) zone allocation granularity, (ii) zone geometry, (iii) write order, and (iv) zone mapping and management strategy are the four main causes behind this. To provide a predictable storage device, we propose SilentZNS, a new holistic zone management approach that expands the design space of zones and allocates blocks to zones on the fly, while minimizing wear, maintaining parallelism, and avoiding superfluous writes to the device. SilentZNS is a flexible zone allocation scheme that departs from traditional logical-to-physical zone mapping and allows arbitrary collections of blocks to be assigned to a zone. SilentZNS further guarantees wear-leveling and competitive read performance, while substantially reducing DLWA. We implement SilentZNS using the state-of-the-art ConfZNS++ emulator and evaluate it on synthetic microbenchmarks and key-value storage engines. We show that SilentZNS reduces superfluous writes, leading to lower DLWA (92% less at 10% zone occupancy), less overall wear (up to 12%), and up to 3.7x faster workload execution.

cs.AR↗

arXiv2025

Reconstructing Sparticle masses at the LHC using Generative Machine Learning

We explore a generative model framework to infer the masses of heavy particles from detector-level data over a broad parameter space. Our model combines a transformer-based detector encoder and a diffusion neural network. We first apply our model to a new physics scenario involving the pair production of wino-like chargino-neutralino, $pp \to \tilde\chi_1^{\pm} \tilde\chi_2^0$, in the $1\ell + 2\gamma + jets$ channel at the high luminosity LHC~(HL-LHC). We find that our framework can achieve mass reconstruction efficiency of $\gtrsim 70\%$ for the lightest neutralino $\tilde\chi_1^0$ and $\gtrsim 40\%$ for the second lightest neutralino $\tilde\chi_2^0$, for a mass tolerance of $\Delta m = 30~$GeV, across the entire parameter space accessible at the HL-LHC. We further extend our analysis to a different scenario with $pp\to\tilde\chi_1^{\pm}\tilde\chi_1^{\mp}+\tilde\chi_1^{\pm}\tilde\chi_2^0$ pair production at the HL-LHC in the $4\ell+\rm E{\!\!\!/}_T$ channel, and for a fixed value of $m_{\tilde\chi_2^0}$, we obtain reconstruction efficiencies $\gtrsim80\%$ over a wide range of $m_{\tilde\chi_1^0}$ for $\Delta m = 30~$GeV.

hep-ph↗

arXiv2025

commensurability: a Python package for classifying astronomical orbits based on their toroid volume

As a star orbits the center of its host galaxy, the trajectory is encompassed within a 3D toroid. The orbit probes all points in this toroid, unless its orbital frequencies exhibit integer ratios (commensurate frequencies), in which case a small sub-volume is traversed. commensurability is a Python package that implements a tessellation-based algorithm for identifying orbital families that satisfy commensurabilities by measuring the toroid volume traversed over orbit integration. Compared to standard orbit classification methods such as frequency analysis, tessellation analysis relies on configuration space properties alone, making classification results more robust to frequency instabilities or limited integration times. The package provides a framework for analyzing phase-space coordinates using tessellation analysis, including a subpackage for the implementation of the general tessellation algorithm. The package is to be used with a galactic dynamics library; it currently supports AGAMA, gala, and galpy.

astro-ph.GA↗

arXiv2024

Searches for the BSM scenarios at the LHC using decision tree based machine learning algorithms: A comparative study and review of Random Forest, Adaboost, XGboost and LightGBM frameworks

Machine learning algorithms are now being extensively used in our daily lives, spanning across diverse industries as well as academia. In the field of high energy physics (HEP), the most common and challenging task is separating a rare signal from a much larger background. The boosted decision tree (BDT) algorithm has been a cornerstone of the high energy physics for analyzing event triggering, particle identification, jet tagging, object reconstruction, event classification, and other related tasks for quite some time. This article presents a comprehensive overview of research conducted by both HEP experimental and phenomenological groups that utilize decision tree algorithms in the context of the Standard Model and Supersymmetry (SUSY). We also summarize the basic concept of machine learning and decision tree algorithm along with the working principle of \texttt{Random Forest}, \texttt{AdaBoost} and two gradient boosting frameworks, such as \texttt{XGBoost}, and \texttt{LightGBM}. Using a case study of electroweakino productions at the high luminosity LHC, we demonstrate how these algorithms lead to improvement in the search sensitivity compared to traditional cut-based methods in both compressed and non-compressed R-parity conserving SUSY scenarios. The effect of different hyperparameters and their optimization, feature importance study using SHapley values are also discussed in detail.

hep-ph↗

arXiv2023

Slepton searches in the trilinear RPV SUSY scenarios at the HL-LHC and HE-LHC

In this work we have studied a multi-lepton final state arising from sneutrino and left-handed slepton production at the high luminosity and high energy LHC in the context of R-parity violating supersymmetry when only the lepton number violating $\lambda_{121}$ and/or $\lambda_{122}$ couplings are non-zero. We have taken into account both pair production and associated production of the three generations of left-handed sleptons and sneutrinos, which are assumed to be mass degenerate. The lightest supersymmetric particle is assumed to be bino and it decays via the R-parity violating couplings into light leptons and neutrinos. Our final state has a large lepton multiplicity, $N_{l}\geq 4~(l=e,~\mu)$. We perform both cut-based and machine learning based analyses for comparison. We present our results in the bino-slepton/sneutrino mass plane in terms of exclusion and discovery reach at the LHC. Following our analysis, the slepton mass can be discovered upto $\sim$ 1.54 TeV and excluded upto $\sim$ 1.87 TeV at the high luminosity LHC while these ranges go upto $\sim$ 2.46 TeV and $\sim$ 3.06 TeV respectively at the high energy LHC.

hep-ph↗

arXiv2023

Improving sensitivity of trilinear RPV SUSY searches using machine learning at the LHC

In this work, we have explored the sensitivity of multilepton final states in probing the gaugino sector of R-parity violating supersymmetric scenario with specific lepton number violating trilinear couplings ($\lambda_{ijk}$) being non-zero. The gaugino spectrum is such that the charged leptons in the final state can arise from the R-parity violating decays of the lightest supersymmetric particle (LSP) as well as R-parity conserving decays of the next-to-LSP (NLSP). Apart from a detailed cut-based analysis, we have also performed a machine learning-based analysis using boosted decision tree algorithm which provides much better sensitivity. In the scenarios with non-zero $\lambda_{121}$ and/or $\lambda_{122}$ couplings, the LSP pair in the final states decays to $4l~(l = e, \mu) + \rm E{\!\!\!/}_T$ final states with $100\%$ branching ratio. We have shown that under this circumstance, a final state with $\ge 4l$ has the highest sensitivity in probing the gaugino masses. We also discuss how the sensitivity can change in the presence of $\tau$ lepton(s) in the final state due to other choices of trilinear couplings. We present our results through the estimation of the discovery and exclusion contours in the gaugino mass plane for both the HL-LHC and the HE-LHC. For $\lambda_{121}$ and/or $\lambda_{122}$ nonzero scenario, the projected 2$\sigma$ exclusion limit on NLSP masses reaches upto 2.37 TeV and 4 TeV for the HL-LHC and the HE-LHC respectively by using a machine learning based algorithm. We obtain an enhancement of $\sim$ 380 (190) GeV in the projected 2$\sigma$ exclusion limit on the NLSP masses at the 27 (14) TeV LHC. Considering the same final state ($N_l \geq 4$) for $\lambda_{133}$ and/or $\lambda_{233}$ non-zero scenario, we find that the corresponding 2$\sigma$ projected limits are $\sim$ 1.97 TeV and $\sim$ 3.25 TeV for the HL-LHC and HE-LHC respectively.

hep-ph↗

arXiv2022

Sensitivity mapping of TBL wall-pressure spectra with CFD turbulence models for wind tunnel test result prediction

In the present work, an attempt is made to map the sensitivity of the existing zero pressure gradient (ZPG) turbulent boundary layer (TBL) wall-pressure spectrum models with different TBL parameters, and eventually, with different Reynolds Averaged Navier Stokes (RANS) turbulence models, simulated in OpenFOAM and ANSYS Fluent solvers. This study will help future researchers to choose a particular RANS turbulence model vis-à-vis a particular wall-spectrum model in order to obtain a reasonably accurate wind tunnel result predicting capability. First, the best-predicting pressure spectrum models are selected by comparing them with wind tunnel test data. Next, considering the experimental TBL parameters as benchmarks, errors in RANS-produced data are estimated. Furthermore, wall-pressure spectra are calculated following semi-empirical spectrum models using TBL parameter feed obtained from experiments and computational fluid dynamics (CFD) simulations. Finally, sensitivity mapping is performed between spectrum models and the RANS models, with different normalized wall-normal distances (y+).

physics.flu-dyn↗

arXiv2022

Constructing and Analyzing the LSM Compaction Design Space (Updated Version)

Log-structured merge (LSM) trees offer efficient ingestion by appending incoming data, and thus, are widely used as the storage layer of production NoSQL data stores. To enable competitive read performance, LSM-trees periodically re-organize data to form a tree with levels of exponentially increasing capacity, through iterative compactions. Compactions fundamentally influence the performance of an LSM-engine in terms of write amplification, write throughput, point and range lookup performance, space amplification, and delete performance. Hence, choosing the appropriate compaction strategy is crucial and, at the same time, hard as the LSM-compaction design space is vast, largely unexplored, and has not been formally defined in the literature. As a result, most LSM-based engines use a fixed compaction strategy, typically hand-picked by an engineer, which decides how and when to compact data. In this paper, we present the design space of LSM-compactions, and evaluate state-of-the-art compaction strategies with respect to key performance metrics. Toward this goal, our first contribution is to introduce a set of four design primitives that can formally define any compaction strategy: (i) the compaction trigger, (ii) the data layout, (iii) the compaction granularity, and (iv) the data movement policy. Together, these primitives can synthesize both existing and completely new compaction strategies. Our second contribution is to experimentally analyze 10 compaction strategies. We present 12 observations and 7 high-level takeaway messages, which show how LSM systems can navigate the compaction design space.

cs.DB↗

arXiv2022

OSM-tree: A Sortedness-Aware Index

Indexes facilitate efficient querying when the selection predicate is on an indexed key. As a result, when loading data, if we anticipate future selective (point or range) queries, we typically maintain an index that is gradually populated as new data is ingested. In that respect, indexing can be perceived as the process of adding structure to an incoming, otherwise unsorted, data collection. The process of adding structure comes at a cost, as instead of simply appending incoming data, every new entry is inserted into the index. If the data ingestion order matches the indexed attribute order, the ingestion cost is entirely redundant and can be avoided (e.g., via bulk loading in a B+-tree). However, state-of-the-art index designs do not benefit when data is ingested in an order that is close to being sorted but not fully sorted. In this paper, we study how indexes can benefit from partial data sortedness or near-sortedness, and we propose an ensemble of techniques that combine bulk loading, index appends, variable node fill/split factor, and buffering, to optimize the ingestion cost of a tree index in presence of partial data sortedness. We further augment the proposed design with necessary metadata structures to ensure competitive read performance. We apply the proposed design paradigm on a state-of-the-art B+-tree, and we propose the Ordered Sort-Merge tree (OSM-tree). OSM-tree outperforms the state of the art by up to 8.8x in ingestion performance in the presence of sortedness, while falling back to a B+-tree's ingestion performance when data is scrambled. OSM-tree offers competitive query performance, leading to performance benefits between 28% and 5x for mixed read/write workloads.

cs.DB↗

arXiv2020

Lethe: A Tunable Delete-Aware LSM Engine (Updated Version)

Data-intensive applications fueled the evolution of log structured merge (LSM) based key-value engines that employ the out-of-place paradigm to support high ingestion rates with low read/write interference. These benefits, however, come at the cost of treating deletes as a second-class citizen. A delete inserts a tombstone that invalidates older instances of the deleted key. State-of-the-art LSM engines do not provide guarantees as to how fast a tombstone will propagate to persist the deletion. Further, LSM engines only support deletion on the sort key. To delete on another attribute (e.g., timestamp), the entire tree is read and re-written. We highlight that fast persistent deletion without affecting read performance is key to support: (i) streaming systems operating on a window of data, (ii) privacy with latency guarantees on the right-to-be-forgotten, and (iii) en masse cloud deployment of data systems that makes storage a precious resource. To address these challenges, in this paper, we build a new key-value storage engine, Lethe, that uses a very small amount of additional metadata, a set of new delete-aware compaction policies, and a new physical data layout that weaves the sort and the delete key order. We show that Lethe supports any user-defined threshold for the delete persistence latency offering higher read throughput ($1.17-1.4\times$) and lower space amplification ($2.1-9.8\times$), with a modest increase in write amplification (between $4\%$ and $25\%$). In addition, Lethe supports efficient range deletes on a secondary delete key by dropping entire data pages without sacrificing read performance nor employing a costly full tree merge.

cs.DB↗