arXiv subjects

Amit Chakraborty

Publications and source records attributed to Amit Chakraborty.

At least 55 records · Page 3Linked to original sources

arXiv2020

Dissipative SymODEN: Encoding Hamiltonian Dynamics with Dissipation and Control into Deep Learning

In this work, we introduce Dissipative SymODEN, a deep learning architecture which can infer the dynamics of a physical system with dissipation from observed state trajectories. To improve prediction accuracy while reducing network size, Dissipative SymODEN encodes the port-Hamiltonian dynamics with energy dissipation and external input into the design of its computation graph and learns the dynamics in a structured way. The learned model, by revealing key aspects of the system, such as the inertia, dissipation, and potential energy, paves the way for energy-based controllers.

cs.LG↗

arXiv2020

Interpretable Deep Learning for Two-Prong Jet Classification with Jet Spectra

Classification of jets with deep learning has gained significant attention in recent times. However, the performance of deep neural networks is often achieved at the cost of interpretability. Here we propose an interpretable network trained on the jet spectrum $S_{2}(R)$ which is a two-point correlation function of the jet constituents. The spectrum can be derived from a functional Taylor series of an arbitrary jet classifier function of energy flows. An interpretable network can be obtained by truncating the series. The intermediate feature of the network is an infrared and collinear safe C-correlator which allows us to estimate the importance of a $S_{2}(R)$ deposit at an angular scale R in the classification. The performance of the architecture is comparable to that of a convolutional neural network (CNN) trained on jet images, although the number of inputs and complexity of architecture is significantly simpler than the CNN classifier. We consider two examples: one is the classification of two-prong jets which differ in color charge of the mother particle, and the other is a comparison between Pythia 8 and Herwig 7 generated jets.

hep-ph↗

arXiv2019

A Conditional Generative Model for Predicting Material Microstructures from Processing Methods

Microstructures of a material form the bridge linking processing conditions - which can be controlled, to the material property - which is the primary interest in engineering applications. Thus a critical task in material design is establishing the processing-structure relationship, which requires domain expertise and techniques that can model the high-dimensional material microstructure. This work proposes a deep learning based approach that models the processing-structure relationship as a conditional image synthesis problem. In particular, we develop an auxiliary classifier Wasserstein GAN with gradient penalty (ACWGAN-GP) to synthesize microstructures under a given processing condition. This approach is free of feature engineering, requires modest domain knowledge and is applicable to a wide range of material systems. We demonstrate this approach using the ultra high carbon steel (UHCS) database, where each microstructure is annotated with a label describing the cooling method it was subjected to. Our results show that ACWGAN-GP can synthesize high-quality multiphase microstructures for a given cooling method.

eess.IV↗

arXiv2018

Flavour-violating decays of mixed top-charm squarks at the LHC

We explore signatures related to squark decays in the framework of non-minimally flavour-violating Supersymmetry. We consider a simplified model where the lightest squark consists of an admixture of charm and top flavour. By recasting the existing LHC searches for top and charm squarks, we show that the limits on squark masses from these analyses are significantly weakened when the top-charm mixing is sizeable. We propose a dedicated search for squarks based on the $tc+{E_{\mathrm{T}}^{\mathrm{miss}}}$ final state which enhances the experimental sensitivity for the case of high mixing, and we map its expected reach for the forthcoming runs of the LHC. We emphasize the role of analyses requiring a jet tagged as produced by the fragmentation of a charm quark in understanding the squark mixing pattern, thus providing a novel handle on new physics. Our results show that, in order to achieve full coverage of the parameter space of supersymmetric models, it is necessary to extend current experimental search programmes with analyses specifically targeting the cases where the lightest top-partner is a mixed state.

hep-ph↗

arXiv2018

Monojet Signatures from Heavy Colored Particles: Future Collider Sensitivities and Theoretical Uncertainties

In models with colored particle $\mathcal{Q}$ that can decay into a dark matter candidate $X$, the relevant collider process $pp\to \mathcal{Q}\bar{\mathcal{Q}}\rightarrow X\bar{X}+$jets gives rise to events with significant transverse momentum imbalance. When the masses of $\mathcal{Q}$ and $X$ are very close, the relevant signature becomes monojet-like, and Large Hadron Collider (LHC) search limits become much less constraining. In this paper, we study the current and anticipated experimental sensitivity to such particles at the High-Luminosity LHC at $\sqrt{s}=14\,\mathrm{TeV}$ with $\mathcal{L}=3\,\mathrm{ab}^{-1}$ of data and the proposed High-Energy LHC at $\sqrt{s}=27\,\mathrm{TeV}$ with $\mathcal{L}=15\,\mathrm{ab}^{-1}$ of data. We estimate the reach for various Lorentz and QCD color representations of $\mathcal{Q}$. Identifying the nature of $\mathcal{Q}$ is very important to understanding the physics behind the monojet signature. Therefore, we also study the dependence of the observables built from the $pp\to\mathcal{Q}\bar{\mathcal{Q}} + j $ process on $\mathcal{Q}$ itself. Using the state-of-the-art Monte Carlo suites MadGraph5_aMC@NLO+Pythia8 and Sherpa, we find that when these observables are calculated at NLO in QCD with parton shower matching and multijet merging, the residual theoretical uncertainties are comparable to differences observed when varying the quantum numbers of $\mathcal{Q}$ itself. We find, however, that the precision achievable with NNLO calculations, where available, can resolve this dilemma.

hep-ph↗

arXiv2018

A Framework for Finding Anomalous Objects at the LHC

Search for new physics events at the LHC mostly rely on the assumption that the events are characterized in terms of standard-reconstructed objects such as isolated photons, leptons, and jets initiated by QCD-partons. While such strategy works for a vast majority of physics beyond the standard model scenarios, there are examples aplenty where new physics give rise to anomalous objects (such as collimated and equally energetic particles, decays due to long lived particles etc.) in the detectors, which can not be classified as any of the standard-objects. Varied methods and search strategies have been proposed, each of which is trained and optimized for specific models, topologies, and model parameters. Further, as LHC keeps excluding all expected candidates for new physics, the need for a generic method/tool that is capable of finding the unexpected can not be understated. In this paper, we propose one such method that relies on the philosophy that all anomalous objects are $\it{not}$ standard-objects. The anomaly finder, we suggest, simply is a collection of vetoes that eliminate all standard-objects up to a pre-determined acceptance rate. Any event containing at least one anomalous object (that passes all these vetoes), can be identified as a candidate for new physics. Subsequent offline analyses can determine the nature of the anomalous object as well as of the event, paving a robust way to search for these new physics scenarios in a model-independent fashion. Further, since the method relies on learning only the standard-objects, for which control samples are readily available from data, one can build the analysis in an entirely data-driven way.

hep-ph↗

arXiv2018

Convolutional Neural Knowledge Graph Learning

Previous models for learning entity and relationship embeddings of knowledge graphs such as TransE, TransH, and TransR aim to explore new links based on learned representations. However, these models interpret relationships as simple translations on entity embeddings. In this paper, we try to learn more complex connections between entities and relationships. In particular, we use a Convolutional Neural Network (CNN) to learn entity and relationship representations in knowledge graphs. In our model, we treat entities and relationships as one-dimensional numerical sequences with the same length. After that, we combine each triplet of head, relationship, and tail together as a matrix with height 3. CNN is applied to the triplets to get confidence scores. Positive and manually corrupted negative triplets are used to train the embeddings and the CNN model simultaneously. Experimental results on public benchmark datasets show that the proposed model outperforms state-of-the-art models on exploring unseen relationships, which proves that CNN is effective to learn complex interactive patterns between entities and relationships.

cs.LG↗

arXiv2017

InverseNet: Solving Inverse Problems with Splitting Networks

We propose a new method that uses deep learning techniques to solve the inverse problems. The inverse problem is cast in the form of learning an end-to-end mapping from observed data to the ground-truth. Inspired by the splitting strategy widely used in regularized iterative algorithm to tackle inverse problems, the mapping is decomposed into two networks, with one handling the inversion of the physical forward model associated with the data term and one handling the denoising of the output from the former network, i.e., the inverted version, associated with the prior/regularization term. The two networks are trained jointly to learn the end-to-end mapping, getting rid of a two-step training. The training is annealing as the intermediate variable between these two networks bridges the gap between the input (the degraded version of output) and output and progressively approaches to the ground-truth. The proposed network, referred to as InverseNet, is flexible in the sense that most of the existing end-to-end network structure can be leveraged in the first network and most of the existing denoising network structure can be used in the second one. Extensive experiments on both synthetic data and real datasets on the tasks, motion deblurring, super-resolution, and colorization, demonstrate the efficiency and accuracy of the proposed method compared with other image processing algorithms.

cs.CV↗

arXiv2017

Search for bottom squarks in the baryon-number violating MSSM

We consider a scenario of Minimal Supersymmetric Standard Model (MSSM) with $R$-parity violation, where the lightest supersymmetric particle (LSP) is the lighter sbottom $(\tilde b_1)$. We study the production of a sbottom pair at the LHC and their subsequent decays through the baryon number violating $(U D D)$ operators leading to a top pair with two light quarks. Looking for both semi-leptonic and fully hadronic (no leptons) final states, we perform cut-based as well as multivariate analyses (MVA) to estimate the signal significance at the 13 TeV run of the LHC. We find that a cut-based analysis can probe sbottom mass upto $\sim 650 $ GeV which may be extended upto $\sim 750 $ GeV using MVA with $300~{\rm fb}^{-1}$ integrated luminosity. The fully hadronic final state, however, is not as promising.

hep-ph↗

arXiv2017

Mixed Higgs-Radion States at the LHC -- a Detailed Study

Light radions constitute one of the few surviving possibilities for observable new particle states at the sub-TeV level which arise in models with extra spacetime dimensions. It is already known that the 125 GeV state discovered at CERN is unlikely to be a pure radion state, since its decays resemble those of the Standard Model Higgs boson too closely. However, due to experimental errors in the measured decay widths, the possibility still remains that it could be a mixture of the radion with one (or more) Higgs states. We use the existing LHC data at 8 and 13 TeV to make a thorough investigation of this possibility. Not surprisingly, it turns out that this model is already constrained quite effectively by direct LHC searches for an additional scalar heavier than 125 GeV. We then make a detailed study of the so-called 'conformal point', where this heavy state practically decouples from (most of) the Standard Model fields. Some projections for the future are also included.

hep-ph↗

arXiv2016

Study of MSSM heavy Higgs bosons decaying into charginos and neutralinos

A multitude of searches have already been performed by the ATLAS and CMS collaborations at the LHC to probe the heavy Higgses of the Minimal Supersymmetric Standard Model (MSSM) through their decay to the Standard Model particles. In this paper, we study the decay of the MSSM heavy Higgses into neutralino and chargino pairs and estimate the maximum possible branching ratios for these `ino' modes being consistent with the present LHC data. After performing a random scan of the relevant electroweakino parameters, we impose the SM 125 GeV Higgs constraints, low energy flavour data as well as current bounds on heavy Higgses from the LHC run-I and run-II data. The present limits on the electroweakino masses and couplings are also considered in our analysis. We choose a few representative benchmark points satisfying all the above-mentioned constraints, and then perform a detailed collider simulation, including fast detector effects, and analyze all the potential SM backgrounds in order to estimate the discovery reach of these heavy Higgses at the LHC. We restrict ourselves within the leptonic cascade decay modes of these heavy Higgses and study the mono-X + MET (X=W,Z) and trilepton + MET signatures in the context of high luminosity run of the 14 TeV LHC.

hep-ph↗

arXiv2016

Chasing New Physics in Stacks of Soft Tracks

In this letter we introduce a new variable $ξ$, namely the number of tracks associated with the primary vertex, which are not parts of reconstructed objects such as jets/isolated leptons etc. We demonstrate its usefulness in the context of new physics searches in the channel monojet$+$missing transverse momentum (MET). In models such as in compressed supersymmetry, events are often characterized by a rather large number of soft partons from the cascade decays, none of which result in reconstructed objects. We find that $ξ$, binned in $p_T$, can discriminate these new physics events from events due to $Z+\text{jets}$, that is, the main background in the channel monojet+ MET. The information contained in soft tracks is largely uncorrelated with traditional variables such as the effective mass, MET, $p_T$ of the jet, etc. and, therefore, can be combined with these to increase the discovery potential by more than $200\%$ (depending on the spectra, of course). In fact, we find that simple cuts on $ξ(p_T)$ along with cuts on MET, and the effective mass outperforms sophisticated optimized MultiVariate Analyses using all conventional variables. One can model the background distribution of $ξ(p_T)$ in an entirely data-driven way, and make these robust against pile-up by identifying the primary vertex.

hep-ph↗

arXiv2016

Probing $(g-2)_μ$ at the LHC in the paradigm of R-parity violating MSSM

The measurement of the anomalous magnetic moment of the muon exhibits a long standing discrepancy compared to the Standard model prediction. In this paper, we concentrate on this issue in the framework of $R$-parity violating Minimal supersymmetric standard model. Such a scenario provides substantial contribution to the anomalous magnetic moment of the muon while satisfying constraints from low energy experimental observables as well as neutrino mass. In addition, we point out that the implication of such operators satisfying muon $g-2$ are immense from the perspective of the LHC experiment, leading to a spectacular four muon final state. We propose an analysis in this particular channel which might help to settle the debate of R-parity violation as a probable explanation for $(g-2)_μ$.

hep-ph↗

arXiv2016

Status of the 98 - 125 GeV Higgs scenario with updated LHC-8 data

In the context of minimal supersymmetric standard model (MSSM), we discuss the possibility of the lightest Higgs boson with mass $M_h = 98 $ GeV to be consistent with the $2.3σ$ excess observed at the LEP in the decay mode $e^+e^- \to Zh$, with $h \to b {\bar b}$. In the same region of the MSSM parameter space, the heavier Higgs boson $(H)$ with mass $M_H \sim 125 $ GeV is required to be consistent with the latest data on Higgs coupling measurements at the end of 7 + 8 TeV LHC run with $25{\rm fb}^{-1}$ of data. While scanning the MSSM parameter space, we impose constraints coming from flavour physics, relic density of the cold dark matter as well as direct dark matter searches. We study the possibility of observing this light Higgs boson in vector boson fusion process and associated production with $W/Z$ boson at the high luminosity $(3000~{\rm fb}^{-1})$ run of the 14 TeV LHC. Our analysis shows that this scenario can hardly be ruled out even at the high luminosity run of the LHC. However, the precise measurement of the Higgs signal strength ratios can play a major role to distinguish this scenario from the canonical MSSM one.

hep-ph↗

arXiv2016

Diphoton resonance at 750 GeV in the broken MRSSM

Non-observation of superpartners of the Standard Model particles at the early runs of the LHC provide strong motivation for an $R$-symmetric minimal supersymmetric Standard Model, or MRSSM. This model also comes with a pair of extra scalars which couple only to superpartners at the tree level. We demonstrate that in the limit when the $U(1)_R$ symmetry is broken, one of these scalars develops all the properties necessary to explain the 750 GeV diphoton resonance recently observed at the LHC, as well as the non-observation of associated signals in other channels. Some confirmatory tests in the upcoming LHC runs are proposed.

hep-ph↗

arXiv2016

Radion Candidate for the LHC Diphoton Resonance

The recent observation of a modest excess in diphoton final states at the LHC, by both the ATLAS and CMS Collaborations, has sparked off the expected race among theorists to find the right explanation for this proto-resonance, assuming that the signal will survive and not prove to be yet another statistical fluctuation. We carry out a general analysis of this `signal' in the case of a scalar which couples only to pairs of gluons (for production) and photons (for diphoton decay modes), and establish that an explanation of the observed resonance, taken together with the null results of new physics searches in all the other channels, requires a scalar with rather exotic behaviour. We then demonstrate that a fairly simple-minded extension of the minimal Randall-Sundrum model can yield a radion candidate which might reproduce this exotic behaviour.

hep-ph↗

arXiv2015

Alternating direction method of multipliers for regularized multiclass support vector machines

The support vector machine (SVM) was originally designed for binary classifications. A lot of effort has been put to generalize the binary SVM to multiclass SVM (MSVM) which are more complex problems. Initially, MSVMs were solved by considering their dual formulations which are quadratic programs and can be solved by standard second-order methods. However, the duals of MSVMs with regularizers are usually more difficult to formulate and computationally very expensive to solve. This paper focuses on several regularized MSVMs and extends the alternating direction method of multiplier (ADMM) to these MSVMs. Using a splitting technique, all considered MSVMs are written as two-block convex programs, for which the ADMM has global convergence guarantees. Numerical experiments on synthetic and real data demonstrate the high efficiency and accuracy of our algorithms.

stat.ML↗

arXiv2015

Proximal gradient method for huberized support vector machine

The Support Vector Machine (SVM) has been used in a wide variety of classification problems. The original SVM uses the hinge loss function, which is non-differentiable and makes the problem difficult to solve in particular for regularized SVMs, such as with $\ell_1$-regularization. This paper considers the Huberized SVM (HSVM), which uses a differentiable approximation of the hinge loss function. We first explore the use of the Proximal Gradient (PG) method to solving binary-class HSVM (B-HSVM) and then generalize it to multi-class HSVM (M-HSVM). Under strong convexity assumptions, we show that our algorithm converges linearly. In addition, we give a finite convergence result about the support of the solution, based on which we further accelerate the algorithm by a two-stage method. We present extensive numerical experiments on both synthetic and real datasets which demonstrate the superiority of our methods over some state-of-the-art methods for both binary- and multi-class SVMs.

stat.ML↗