AGAnushree Ghosh

Anushree Ghosh

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Anushree Ghosh has a total of 387 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Inferring Convolutional Neural Networks' accuracies from their architectural characterizations

    by Duc Hoang, Jesse Hamer, Gabriel N. Perdue, Steven R. Young, Jonathan Miller, Anushree Ghosh

    Convolutional Neural Networks (CNNs) have shown strong promise for analyzing scientific data from many domains including particle imaging detectors. However, the challenge of choosing the appropriate network architecture (depth, kernel shapes, activation functions, etc.) for specific applications and different data sets is still poorly understood. In this paper, we study the relationships between a CNN's architecture and its performance by proposing a systematic language that is useful for comparison between different CNN's architectures before training time. We characterize CNN's architecture by different attributes, and demonstrate that the attributes can be predictive of the networks' performance in two specific computer vision-based physics problems -- event vertex finding and hadron multiplicity classification in the MINERvA experiment at Fermi National Accelerator Laboratory. In doing so, we extract several architectural attributes from optimized networks' architecture for the physics problems, which are outputs of a model selection algorithm called Multi-node Evolutionary Neural Networks for Deep Learning (MENNDL). We use machine learning models to predict whether a network can perform better than a certain threshold accuracy before training. The models perform 16-20% better than random guessing. Additionally, we found an coefficient of determination of 0.966 for an Ordinary Least Squares model in a regression on accuracy over a large population of networks.

  • Open Access

    Indirect Searches for Dark Matter Signatures at INO

    by S Choubey, A Ghosh, D Tiwari

    Neutrino fluxes could arise due to annihilation of Weakly Interactive Massive Particles (WIMPs) in the center of the sun. We study the prospects of search for muon events due to such neutrinos at the upcoming Iron CALorimeter (ICAL) detector to be housed at India-based Neutrino Observatory (INO). Although the atmospheric neutrinos will pose a serious background to the signal neutrinos produced through WIMP annihilation, the former could be supressed significantly by using the directional property of signal neutrinos. For 50kt×10 years of ICAL running and WIMP masses (mχ ) between 3-100 GeV, we perform a χ2 analysis and present expected exclusion regions in the σβख़ — mx and σβι — mx plane, where σβख़ and σβι are the WIMP-nucleon Spin-Dependent (SD) and Spin-Independent (SI) scattering cross-sections, respectively. For mx = 25 GeV, the expected 90 % C.L. exclusion limit on σβख़ are σβॄ < 7.82 x 10−41 cm2 for τ+τ− channel and σβख़ < 1.23 10−39 cm2 for 66 channel. For same m× , the expected 90 % C.L. exclusion limits on σβι are σβι < 8.97 × 10−43 cm2 for τ +τ − channel and σβι < 1.43×10−41 cm2 for 66 channel.

  • Open Access

    Inferring Convolutional Neural Networks' Accuracies from Their Architectural Characterizations

    by Duc Hoang, Jesse Hamer, Gabriel Perdue, Steven Young, Jonathan Miller, Anushree Ghosh

    The challenge of choosing an appropriate convolutional neural network (CNN) architecture for specific applications and different data sets is still poorly understood in the literature. This is problematic, since CNNs have shown strong promise for analyzing scientific data from many domains including particle imaging detectors. In this paper, we proposed a systematic language that is useful for comparison between different CNN's architectures before training time. This helps us predict whether a network can perform better than a certain threshold accuracy before training up to 70% accuracy using simple machine learning models. Additionally, we found a coefficient of determination of 0.966 for an Ordinary Least Squares model in a regression task to predict accuracy of a large population of networks.


  • Open Access

    Tuning the genie pion production model with MINERvA data

    by P. Stowell, L. Pickering, C. Wilkinson, C. Wret, F. Akbar, D. A. Andrade, M. V. Ascencio, L. Bellantoni, A. Bercellie, M. Betancourt, et al.

    Faced with unresolved tensions between neutrino interaction measurements at few-GeV neutrino energies, current experiments are forced to accept large systematic uncertainties to cover discrepancies between their data and model predictions. The widely used pion production model in genie is compared to four MINERνA charged current pion production measurements using nuisance. Tunings, i.e., adjustments of model parameters, to help match genie to MINERνA and older bubble chamber data are presented. We find that scattering off nuclear targets as measured in MINERνA is not in good agreement with expectations based upon scattering off nucleon (hydrogen or deuterium) targets in existing bubble chamber data. An additional ad hoc correction for the low−Q2 region, where collective nuclear effects are expected to be large, is presented. While these tunings and corrections improve the agreement of genie with the data, the modeling is imperfect. The development of these tunings within the nuisance framework allows for straightforward extensions to other neutrino event generators and models, and allows omitting and including new datasets as they become available.


  • Open Access

    Measurement of ν¯μ charged-current single π− production on hydrocarbon in the few-GeV region using MINERvA

    by T Le, F Akbar, L Aliaga, D Andrade, M Ascencio, A Bashyal, A Bercellie, M Betancourt, A Bodek, J Bonilla, et al.

    The antineutrino scattering channel ν¯μCH→μ+π−X (nucleon(s)) is analyzed in the incident energy range 1.5 to 10 GeV using the MINERvA detector at Fermilab. Differential cross sections are reported as functions of μ+ momentum and production angle, π− kinetic energy and production angle, and antineutrino energy and squared four-momentum transfer. Distribution shapes are generally reproduced by simulations based on the GENIE, NuWro, and GiBUU event generators, however GENIE (GiBUU) overestimates (underestimates) the cross section normalizations by 8% (10%). Comparisons of data with the GENIE-based reference simulation probe conventional treatments of cross sections and pion intranuclear rescattering. The distribution of nontrack vertex energy is used to decompose the signal sample into reaction categories, and cross sections are determined for the exclusive reactions μ+π−n and μ+π−p. A similar treatment applied to the published MINERvA sample ν¯μCH→μ+π0X[nucleon(s)] has determined the μ+π0n cross section, and the latter is used with σ(π−n) and σ(π−p) to carry out an isospin decomposition of ν¯μ-induced CC(π). The ratio of magnitudes and relative phase for isospin amplitudes A3 and A1 thereby obtained are: Rν¯=0.99±0.19 and ϕν¯=93°±7°. Our results are in agreement with bubble chamber measurements made four decades ago.


  • Open Access

    Neutron measurements from antineutrino hydrocarbon reactions

    by M. Elkins, T. Cai, J. Chaves, J. Kleykamp, F. Akbar, L. Albin, L. Aliaga, D. A. Andrade, M. V. Ascencio, A. Bashyal, et al.

    Charged-current antineutrino interactions on a hydrocarbon scintillator in the MINERvA detector are used to study activity from their final-state neutrons. To ensure that most of the neutrons are from the primary interaction, rather than hadronic reinteractions in the detector, the sample is limited to momentum transfers below 0.8  GeV/c. From 16 129 interactions, 15 246 neutral particle candidates are observed. The reference simulation predicts 64% of these candidates are due to neutrons from the antineutrino interaction directly but also overpredicts the number of candidates by 15% overall. This discrepancy is beyond the standard uncertainty estimates for models of neutrino interactions and neutron propagation in the detector. We explore these two aspects of the models using the measured distributions for energy deposition, time of flight, position, and speed. We also use multiplicity distributions to evaluate the presence of a two-nucleon knockout process. These results provide critical new information toward a complete description of the hadronic final state of neutrino interactions, which is vital to neutrino oscillation experiments.


  • Open Access

    Prospects of indirect searches for dark matter annihilations in the earth with ICAL@INO

    by Deepak Tiwari, Sandhya Choubey, Anushree Ghosh

    We study the prospects of detecting muon events at the upcoming Iron CALorimeter (ICAL) detector to be built at the proposed India-based Neutrino Observatory (INO) facility due to neutrinos arising out of annihilation of Weakly Interactive Massive Particles (WIMP) in the centre of the earth. The atmospheric neutrinos coming from the direction of earth core presents an irreducible background. We consider 50kt × 10 years of ICAL running and WIMP masses between 10-100 GeV and present 90% C.L. exclusion sensitivity limits on σ SI which is the WIMP-nucleon Spin Independent (SI) interaction cross-section. The expected sensitivity limits calculated for ICAL for the WIMP annihilation in the earth are more stringent than the limits obtained by any other neutrino detector. For a WIMP mass of 52.14 GeV, where the signal fluxes are enhanced due to resonance capture of WIMP in earth due to Fe nuclei, the sensitivity limits, assuming 100% branching ratio for each channel, are: σ SI = 3.43 × 10−45 cm2 for ν μ $$ {\overline{\nu}}_{\mu } $$ ν ¯ μ channel, σ SI = 1.02 × 10−44 cm2 for τ + τ − channel and σ SI = 5.36 × 10−44 cm2 for $$ b\;\overline{b} $$ b b ¯ channel.


  • Open Access

    Measurement of quasielastic-like neutrino scattering at ⟨Eν⟩∼3.5  GeV on a hydrocarbon target

    by D Ruterbories, K Hurtado, J Osta, F Akbar, L Aliaga, D Andrade, M Ascencio, A Bashyal, A Bercellie, M Betancourt, et al.

    MINERvA presents a new analysis of neutrino induced quasielastic-like interactions in a hydrocarbon tracking target. We report a double-differential cross section using the muon transverse and longitudinal momentum. In addition, differential cross sections as a function of the square of the four-momentum transferred and the neutrino energy are calculated using a quasielastic hypothesis. Finally, an analysis of energy deposited near the interaction vertex is presented. These results are compared to modified genie predictions as well as a NuWro prediction. All results use a data set produced by 3.34×1020 protons on target creating a neutrino beam with a peak energy of approximately 3.5 GeV.


  • Open Access

    Reducing model bias in a deep learning classifier using domain adversarial neural networks in the MINERvA experiment

    by G.N. Perdue, A. Ghosh, M. Wospakrik, F. Akbar, D.A. Andrade, M. Ascencio, L. Bellantoni, A. Bercellie, M. Betancourt, G. F. R. Caceres Vera, et al.


  • Open Access

    Prospects of indirect searches for dark matter annihilations in the earth with ICAL@INO

    by Deepak Tiwari, Sandhya Choubey, Anushree Ghosh

    We study the prospects of detecting muon events at the upcoming Iron CALorimeter (ICAL) detector to be built at the proposed India-based Neutrino Observatory (INO) facility due to neutrinos arising out of annihilation of Weakly Interactive Massive Particles (WIMP) in the centre of the earth. The atmospheric neutrinos coming from the direction of earth core presents an irreducible background. We consider 50kt $\times$ 10 years of ICAL running and WIMP masses between 10-100 GeV and present 90 \% C.L. exclusion sensitivity limits on $\sigma_{SI}$ which is the WIMP-nucleon Spin Independent (SI) interaction cross-section. The expected sensitivity limits calculated for ICAL for the WIMP annihilation in the earth are more stringent than the limits obtained by any other indirect detection experiment. For a WIMP mass of ~$52.14 \textup{ GeV}$, where the signal fluxes are enhanced due to resonance capture of WIMP in earth due to Fe nuclei, the sensitivity limits, assuming 100\% branching ratio for each channel, are : $\sigma_{SI} =1.02\times 10^{-44}~cm^2$ for the $\tau^{+} \tau^{-}$ channel and $\sigma_{SI} =5.36\times 10^{-44} ~cm^2$ for the $b ~\bar{b}$ channel.