arXiv subjects

Vaishnavi Patil

Publications and source records attributed to Vaishnavi Patil.

8 recordsLinked to original sources

arXiv2026

Aspects of Carrollian field theory from holography

We study holography for three-dimensional asymptotically flat spacetimes in which the bulk gravitational dynamics is conjectured to be dual to a two-dimensional Carrollian (equivalently BMS) conformal field theory. Such theories are known to arise as ultrarelativistic ($c\rightarrow0$) contractions of relativistic 2d conformal field theories. Under this contraction, the Virasoro algebra becomes the conformal Carrollian algebra and the Brown-Henneaux central charges become $c_L=0$, $c_M=\frac{3}{G}$. In this article, we recover these central charges directly from holography. We first construct the holographic quasilocal stress tensor for asymptotically flat spacetimes carrying Bondi mass and angular momentum. Under BMS$_3$ transformations, the stress tensor acquires an inhomogeneous term, the ``BMS Schwarzian", and extracting the central charges from it reproduces the algebraic result exactly. We then obtain the action for the boundary BMS Schwarzian modes. Finally, we show that the holographic stress-tensor correlators satisfy the expected Ward identity, while stress-tensor conservation obeys flux-balance laws.

hep-th↗

arXiv2026

Holographic entanglement entropy with conformal boundary conditions

We study holographic entanglement entropy in 3-dimensional AdS gravity with conformal boundary conditions which fix the conformal class of the boundary metric and its extrinsic curvature, $K$, while leaving the Weyl mode dynamical. Extending Lewkowycz-Maldacena-Dong's replica construction, we derive the corresponding holographic entanglement entropy formula. We show that the fluctuating Weyl mode does not contribute additional entropy. The entropy of the full boundary is therefore the Bekenstein-Hawking entropy, and the entropy of a boundary subregion continues to obey the Ryu-Takayanagi prescription, namely, the area of a minimal surface divided by $4G_N$. We carry out explicit calculations for global AdS, rotating and non-rotating BTZ geometries. For an interval in AdS$_3$ we find that the $K$-dependent holographic entanglement entropy is governed by $c_m=\frac{3 \ell}{2 G_N}.$ For a thermal state at high conformal temperatures we find that, the entropy is governed by $c_{\rm eff}=\frac{3 \ell}{2 G_N} \frac{K \ell- \sqrt{K^2 \ell^2-4}}{2}$, the same effective central charge that governs the Cardy-like density of states, in agreement with previous results in the literature. Finally, we also compute the entanglement entropy directly from the conjectured dual boundary theory - a holographic CFT coupled with time-like Liouville theory and deformed by a marginal $T \bar{T}$ like operator - and find $S_{EE} =\frac{c_{\rm eff}}{3} \ln\!\left(\frac{2 R \sinϕ_0}ε\right) .$ This result for the state with no operator insertions (vacuum state ), provides an independent boundary realization of $c_{\mathrm{eff}}$ while clarifying that this state is not the state dual to global AdS.

hep-th↗

arXiv2025

The landscape of complexity measures in 2D gravity

We investigate the broad landscape of holographic complexity measures for theories dual to two-dimensional (2D) dilaton gravity. Previous studies have largely focused on the complexity=volume and complexity=action proposals for holographic complexity. Here we systematically construct and analyze a wide class of generalized complexity functionals, focusing on codimension-one bulk observables. Two complementary approaches are presented: one inspired by dimensional reduction of codimension-one observables from higher-dimensional gravity, and another that adopts a purely 2D perspective. We verify the resulting observables exhibit hallmark features of complexity, such as linear growth at late times and the switchback effect. We further offer heuristic interpretations of the role of multiple extremal surfaces when they appear. Finally, we comment on the bulk-to-boundary dictionary via the covariant Peierls bracket in 2D gravity. Our work lays the groundwork for a richer understanding of quantum complexity in low-dimensional holographic dualities.

hep-th↗

arXiv2025

Detecting and Monitoring Bias for Subgroups in Breast Cancer Detection AI

Automated mammography screening plays an important role in early breast cancer detection. However, current machine learning models, developed on some training datasets, may exhibit performance degradation and bias when deployed in real-world settings. In this paper, we analyze the performance of high-performing AI models on two mammography datasets-the Emory Breast Imaging Dataset (EMBED) and the RSNA 2022 challenge dataset. Specifically, we evaluate how these models perform across different subgroups, defined by six attributes, to detect potential biases using a range of classification metrics. Our analysis identifies certain subgroups that demonstrate notable underperformance, highlighting the need for ongoing monitoring of these subgroups' performance. To address this, we adopt a monitoring method designed to detect performance drifts over time. Upon identifying a drift, this method issues an alert, which can enable timely interventions. This approach not only provides a tool for tracking the performance but also helps ensure that AI models continue to perform effectively across diverse populations.

cs.CV↗

arXiv2023

Lorentzian threads and generalized complexities

Recently, an infinite class of holographic generalized complexities was proposed. These gravitational observables display the behavior required to be duals of complexity, in particular, linear growth at late times and switchback effect. In this work, we aim to understand generalized complexities in the framework of Lorentzian threads. We reformulate the problem in terms of thread distributions and measures and present a program to calculate the infinite family of codimension-one observables. We also outline a path to understand, using threads, the more subtle case of codimension-zero observables.

hep-th↗

arXiv2023

ProtoVAE: Prototypical Networks for Unsupervised Disentanglement

Generative modeling and self-supervised learning have in recent years made great strides towards learning from data in a completely unsupervised way. There is still however an open area of investigation into guiding a neural network to encode the data into representations that are interpretable or explainable. The problem of unsupervised disentanglement is of particular importance as it proposes to discover the different latent factors of variation or semantic concepts from the data alone, without labeled examples, and encode them into structurally disjoint latent representations. Without additional constraints or inductive biases placed in the network, a generative model may learn the data distribution and encode the factors, but not necessarily in a disentangled way. Here, we introduce a novel deep generative VAE-based model, ProtoVAE, that leverages a deep metric learning Prototypical network trained using self-supervision to impose these constraints. The prototypical network constrains the mapping of the representation space to data space to ensure that controlled changes in the representation space are mapped to changes in the factors of variations in the data space. Our model is completely unsupervised and requires no a priori knowledge of the dataset, including the number of factors. We evaluate our proposed model on the benchmark dSprites, 3DShapes, and MPI3D disentanglement datasets, showing state of the art results against previous methods via qualitative traversals in the latent space, as well as quantitative disentanglement metrics. We further qualitatively demonstrate the effectiveness of our model on the real-world CelebA dataset.

cs.LG↗

arXiv2022

DOT-VAE: Disentangling One Factor at a Time

As we enter the era of machine learning characterized by an overabundance of data, discovery, organization, and interpretation of the data in an unsupervised manner becomes a critical need. One promising approach to this endeavour is the problem of Disentanglement, which aims at learning the underlying generative latent factors, called the factors of variation, of the data and encoding them in disjoint latent representations. Recent advances have made efforts to solve this problem for synthetic datasets generated by a fixed set of independent factors of variation. Here, we propose to extend this to real-world datasets with a countable number of factors of variations. We propose a novel framework which augments the latent space of a Variational Autoencoders with a disentangled space and is trained using a Wake-Sleep-inspired two-step algorithm for unsupervised disentanglement. Our network learns to disentangle interpretable, independent factors from the data ``one at a time", and encode it in different dimensions of the disentangled latent space, while making no prior assumptions about the number of factors or their joint distribution. We demonstrate its quantitative and qualitative effectiveness by evaluating the latent representations learned on two synthetic benchmark datasets; DSprites and 3DShapes and on a real datasets CelebA.

cs.LG↗

arXiv2020

Page Curve and the Information Paradox in Flat Space

Asymptotic Causal Diamonds (ACDs) are a natural flat space analogue of AdS causal wedges, and it has been argued previously that they may be useful for understanding bulk locality in flat space holography. In this paper, we use ACD-inspired ideas to argue that there exist natural candidates for Quantum Extremal Surfaces (QES) and entanglement wedges in flat space, anchored to the conformal boundary. When there is a holographic screen at finite radius, we can also associate entanglement wedges and entropies to screen sub-regions, with the system naturally coupled to a sink. The screen and the boundary provide two complementary ways of formulating the information paradox. We explain how they are related and show that in both formulations, the flat space entanglement wedge undergoes a phase transition at the Page time in the background of an evaporating Schwarzschild black hole. Our results closely parallel recent observations in AdS, and reproduce the Page curve. That there is a variation of the argument that can be phrased directly in flat space without reliance on AdS, is a strong indication that entanglement wedge phase transitions may be key to the information paradox in flat space as well. Along the way, we give evidence that the entanglement entropy of an ACD is a well-defined, and likely instructive, quantity. We further note that the picture of the sink we present here may have an understanding in terms of sub-matrix deconfinement in a large-$N$ setting.

hep-th↗