arXiv subjects

Puneet Batra

Publications and source records attributed to Puneet Batra.

13 recordsLinked to original sources

arXiv2023

Latent Space Explorer: Visual Analytics for Multimodal Latent Space Exploration

Machine learning models built on training data with multiple modalities can reveal new insights that are not accessible through unimodal datasets. For example, cardiac magnetic resonance images (MRIs) and electrocardiograms (ECGs) are both known to capture useful information about subjects' cardiovascular health status. A multimodal machine learning model trained from large datasets can potentially predict the onset of heart-related diseases and provide novel medical insights about the cardiovascular system. Despite the potential benefits, it is difficult for medical experts to explore multimodal representation models without visual aids and to test the predictive performance of the models on various subpopulations. To address the challenges, we developed a visual analytics system called Latent Space Explorer. Latent Space Explorer provides interactive visualizations that enable users to explore the multimodal representation of subjects, define subgroups of interest, interactively decode data with different modalities with the selected subjects, and inspect the accuracy of the embedding in downstream prediction tasks. A user study was conducted with medical experts and their feedback provided useful insights into how Latent Space Explorer can help their analysis and possible new direction for further development in the medical domain.

cs.LG↗

arXiv2021

Patient Contrastive Learning: a Performant, Expressive, and Practical Approach to ECG Modeling

Supervised machine learning applications in health care are often limited due to a scarcity of labeled training data. To mitigate this effect of small sample size, we introduce a pre-training approach, Patient Contrastive Learning of Representations (PCLR), which creates latent representations of ECGs from a large number of unlabeled examples. The resulting representations are expressive, performant, and practical across a wide spectrum of clinical tasks. We develop PCLR using a large health care system with over 3.2 million 12-lead ECGs, and demonstrate substantial improvements across multiple new tasks when there are fewer than 5,000 labels. We release our model to extract ECG representations at https://github.com/broadinstitute/ml4h/tree/master/model_zoo/PCLR.

cs.LG↗

arXiv2008

A Composite Twin Higgs Model

Twin Higgs models are economical extensions of the Standard Model that stabilize the electroweak scale. In these theories the Higgs field is a pseudo Nambu-Goldstone boson that is protected against radiative corrections up to scales of order 5 TeV by a discrete parity symmetry. We construct, for the first time, a class of composite twin Higgs models based on confining QCD-like dynamics. These theories naturally incoporate a custodial isospin symmetry and predict a rich spectrum of particles with masses of order a TeV that will be accessible at the LHC.

hep-ph↗

arXiv2008

The Supersymmetric Higgs

In the Minimal Supersymmetric Standard Model (the MSSM), the electroweak symmetry is restored as supersymmetry-breaking terms are turned off. We describe a generic extension of the MSSM where the electroweak symmetry is broken in the supersymmetric limit. We call this limit the "sEWSB" phase, short for supersymmetric electroweak symmetry breaking. We define this phase in an effective field theory that only contains the MSSM degrees of freedom. The sEWSB vacua naturally have an inverted scalar spectrum, where the heaviest CP-even Higgs state has Standard Model-like couplings to the massive vector bosons; experimental constraints in the scalar Higgs sector are more easily satisfied than in the MSSM.

hep-ph↗

arXiv2008

Pseudo-redundant vacuum energy

We discuss models that can account for today's dark energy. The underlying cosmological constant may be Planck scale but starts as a redundant coupling which can be eliminated by a field redefinition. The observed vacuum energy arises when the redundancy is explicitly broken, say by a non-minimal coupling to curvature. We give a recipe for constructing models, including R + 1/R type models, that realize this mechanism and satisfy all solar system constraints on gravity. A similar model, based on Gauss-Bonnet gravity, provides a technically natural explanation for dark energy and exhibits an interesting see-saw behavior: a large underlying cosmological constant gives rise to both low and high curvature solutions. Such models could be statistically favored in the string landscape.

hep-th↗

arXiv2007

Symmetry Breaking Patterns for the Little Higgs from Strong Dynamics

We show how the symmetry breaking pattern of the simplest little Higgs model, and that of the smallest moose model that incorporates an approximate custodial SU(2), can be realized through the condensation of strongly coupled fermions. In each case a custodial SU(2) symmetry of the new strong dynamics limits the sizes of corrections to precision electroweak observables. In the case of the simplest little Higgs, there are no new light states beyond those present in the original model. However, our realization of the symmetry breaking pattern of the moose model predicts an additional scalar field with mass of order a TeV or higher that has exactly the same quantum numbers as the Standard Model Higgs and which decays primarily to third generation quarks.

hep-ph↗

arXiv2007

Transitions Between de Sitter Minima

We investigate the cosmological signatures of instantons mediating tunneling between de Sitter minima. For generic potentials the Coleman-de Luccia instanton does not necessarily exist; when it does not, the instanton which contributes to the decay rate is the trivial constant solution, known as the Hawking-Moss instanton. With the aid of a toy model we interpret this solution and describe the resulting cosmology. In neither the Coleman-de Luccia nor Hawking-Moss case can the resulting cosmology be closed. An observation of significant positive curvature would therefore rule out the possibility that our universe arose from any transition from a neighboring minimum in the string-theory landscape.

hep-th↗

arXiv2006

Measuring the W-t-b Interaction at the ILC

The large top quark mass suggests that the top plays a pivotal role in Electroweak symmetry-breaking dynamics and, as a result, may have modified couplings to Electroweak bosons. Hadron colliders can provide measurements of these couplings at the ~10% level, and one of the early expected triumphs of the International Linear Collider is to reduce these uncertainties to the per cent level. In this article, we propose the first direct measurement of the Standard Model W-t-b coupling at the ILC, from measurements of t tbar-like signals below the t tbar production threshold. We estimate that the ILC with 100 fb^{-1} can measure a combination of the coupling and top width to high precision, and when combined with a direct measurement of the top width from the above-threshold scan, results in a model-independent measurement of the W-t-b interaction of the order of ~ 3%.

hep-ph↗

arXiv2005

Anomaly-Free Sets of Fermions

We present new techniques for finding anomaly-free sets of fermions. Although the anomaly cancellation conditions typically include cubic equations with integer variables that cannot be solved in general, we prove by construction that any chiral set of fermions can be embedded in a larger set of fermions which is chiral and anomaly-free. Applying these techniques to extensions of the Standard Model, we find anomaly-free models that have arbitrary quark and lepton charges under an additional U(1) gauge group.

hep-ph↗

arXiv2004

Perturbative, Non-Supersymmetric Completions of the Little Higgs

The little Higgs mechanism produces a light 100 GeV Higgs while raising the natural cutoff from 1 TeV to 10 TeV. We attempt an iterative little Higgs mechanism to produce multiple factors of 10 between the cutoff and the 100 GeV Higgs mass in a perturbative theory. In the renormalizable sector of the theory, all quantum corrections to the Higgs mass proportional to mass scales greater than 1 TeV are absent -- this includes quadratically divergent, log-divergent, and finite loops at all orders. However, even loops proportional to scales just a factor of 10 above the Higgs (or any other scalar) mass come with large numerical factors that reintroduce fine-tuning. Top loops, for example, produce an expansion parameter of not 1/(4 pi) but 1/5. The geometric increase in the number of fields at higher energies simply exacerbates this problem. We build a complete two-stage model up to 100 TeV, show that direct sensitivity of the electroweak scale to the cutoff is erased, and estimate the tuning due to large numerical factors. We then discuss the possibility, in a toy model with only scalar and gauge fields, of generating a tower of little Higgs theories and show that the theory quickly becomes a large-N gauge theory with ~ N fundamental scalars. We find evidence that at least this toy model could successfully generate light scalars with an exponentially large cutoff in the absence of supersymmetry or strong dynamics. The fine-tuning is not completely eliminated, but evidence suggests that this result is model dependent. We then speculate as to how one might marry a working tower of fields of this type at high scales to a realistic theory at the weak scale.

hep-ph↗

arXiv2004

Running into New Territory in SUSY Parameter Space

The LEP-II bound on the light Higgs mass rules out the vast majority of parameter space left to the Minimal Supersymmetric Standard Model (MSSM) with weak-scale soft-masses. This suggests the importance of exploring extensions of the MSSM with non-minimal Higgs physics. In this article, we explore a theory with an additional singlet superfield and an extended gauge sector. The theory has a number of novel features compared to both the MSSM and Next-to-MSSM, including easily realizing a light CP-even Higgs mass consistent with LEP-II limits, tan(beta) < 1, and a lightest Higgs which is charged. These features are achieved while remaining consistent with perturbative unification and without large stop-masses. Discovery modes at the Tevatron and LHC are discussed.

hep-ph↗

arXiv2004

The Higgs Mass Bound in Gauge Extensions of the Minimal Supersymmetric Standard Model

The minimal supersymmetric standard model, and extensions, have stringent upper bounds on the mass of the lightest Higgs boson if perturbativity up to the Planck scale is assumed. We argue that these bounds are softened tremendously if the Higgs is charged under an asymptotically free gauge group. We present a model with an additional SU(2) gauge group which easily produces Higgs masses above 200 GeV while avoiding electroweak constraints. If one allows some fine-tuning of the high-scale value of the gauge coupling, Higgs masses greater than 350 GeV are achieved. Unification of couplings is predicted to similar accuracy as in the minimal supersymmetric standard model with only small deviations at the two-loop level.

hep-ph↗

arXiv2001

A Determination of the Hubble Constant from Cepheid Distances and a Model of the Local Peculiar Velocity Field

We present a measurement of the Hubble Constant based on Cepheid distances to 27 galaxies within 20 Mpc. We take the Cepheid data from published measurements by the Hubble Telescope Key Project on the Distance Scale (H0KP). We calibrate the Cepheid Period-Luminosity (PL) relation with data from over 700 Cepheids in the LMC obtained by the OGLE collaboration; we assume an LMC distance modulus of 18.50 mag (d=50.1 kpc). Using this PL calibration we obtain new distances to the H0KP galaxies. We correct the redshifts of these galaxies for peculiar velocities using two distinct velocity field models: the phenomenological model of Tonry et al. and a model based on the IRAS density field and linear gravitational instability theory. We combine the Cepheid distances with the corrected redshifts for the 27 galaxies to derive H_0, the Hubble constant. The results are H_0 = 85 +/- 5 km/s/Mpc (random error) at 95% confidence when the IRAS model is used, and 92 +/- 5 km/s/Mpc when the phenomenological model is used. The IRAS model is a better fit to the data and the Hubble constant it returns is more reliable. Systematic error stems mainly from LMC distance uncertainty which is not directly addressed by this paper. Our value of H_0 is significantly larger than that quoted by the H0KP, H_0 = 71 +/- 6 km/s/Mpc. Cepheid recalibration explains ~30% of this difference, velocity field analysis accounts for ~70%. We discuss in detail possible reasons for this discrepancy and future study needed to resolve it.

astro-ph↗