arXiv subjects

Shubham Maheshwari

Publications and source records attributed to Shubham Maheshwari.

9 recordsLinked to original sources

arXiv2025

AUGUSTUS: An LLM-Driven Multimodal Agent System with Contextualized User Memory

Riding on the success of LLMs with retrieval-augmented generation (RAG), there has been a growing interest in augmenting agent systems with external memory databases. However, the existing systems focus on storing text information in their memory, ignoring the importance of multimodal signals. Motivated by the multimodal nature of human memory, we present AUGUSTUS, a multimodal agent system aligned with the ideas of human memory in cognitive science. Technically, our system consists of 4 stages connected in a loop: (i) encode: understanding the inputs; (ii) store in memory: saving important information; (iii) retrieve: searching for relevant context from memory; and (iv) act: perform the task. Unlike existing systems that use vector databases, we propose conceptualizing information into semantic tags and associating the tags with their context to store them in a graph-structured multimodal contextual memory for efficient concept-driven retrieval. Our system outperforms the traditional multimodal RAG approach while being 3.5 times faster for ImageNet classification and outperforming MemGPT on the MSC benchmark.

cs.AI↗

arXiv2021

WiseR: An end-to-end structure learning and deployment framework for causal graphical models

Structure learning offers an expressive, versatile and explainable approach to causal and mechanistic modeling of complex biological data. We present wiseR, an open source application for learning, evaluating and deploying robust causal graphical models using graph neural networks and Bayesian networks. We demonstrate the utility of this application through application on for biomarker discovery in a COVID-19 clinical dataset.

cs.LG↗

arXiv2021

An anisotropic bouncing universe in non-local gravity

We show that it is possible to realize a cosmological bouncing solution in an anisotropic but homogeneous Bianchi-I background in a class of non-local, infinite derivative theories of gravity. We show that the anisotropic shear grows slower than in general relativity during the contraction phase, peaks to a finite value at the bounce point, and then decreases as the universe asymptotes towards isotropy and homogeneity, and ultimately to de Sitter. Along with a cosmological constant, the matter sector required to drive such a bounce is found to consist of three components - radiation, stiff matter and $k$-matter (whose energy density decays like the inverse square of the average scale factor). Generically, $k$-matter exerts anisotropic pressures. We will test the bouncing solution in local and non-local gravity and show that in the latter case it is possible to simultaneously satisfy positivity of energy density and, at least in the late time de Sitter phase, avoid the introduction of propagating ghost/tachyonic modes.

gr-qc↗

arXiv2020

Stable, non-singular bouncing universe with only a scalar mode

In this paper, we study a class of higher derivative, non-local gravity which admits homogeneous and isotropic non-singular, bouncing universes in the absence of matter. At the linearized level, the theory propagates only a scalar degree of freedom, and no vector or tensor modes. The scalar can be made free from perturbative ghost instabilities, and has oscillatory and bounded evolution across the bounce.

gr-qc↗

arXiv2020

Harnessing GANs for Zero-shot Learning of New Classes in Visual Speech Recognition

Visual Speech Recognition (VSR) is the process of recognizing or interpreting speech by watching the lip movements of the speaker. Recent machine learning based approaches model VSR as a classification problem; however, the scarcity of training data leads to error-prone systems with very low accuracies in predicting unseen classes. To solve this problem, we present a novel approach to zero-shot learning by generating new classes using Generative Adversarial Networks (GANs), and show how the addition of unseen class samples increases the accuracy of a VSR system by a significant margin of 27% and allows it to handle speaker-independent out-of-vocabulary phrases. We also show that our models are language agnostic and therefore capable of seamlessly generating, using English training data, videos for a new language (Hindi). To the best of our knowledge, this is the first work to show empirical evidence of the use of GANs for generating training samples of unseen classes in the domain of VSR, hence facilitating zero-shot learning. We make the added videos for new classes publicly available along with our code.

cs.LG↗

arXiv2019

Perturbations in higher derivative gravity beyond maximally symmetric spacetimes

We study (covariant) scalar-vector-tensor (SVT) perturbations of infinite derivative gravity (IDG), at the quadratic level of the action, around conformally-flat, covariantly constant curvature backgrounds which are not maximally symmetric spacetimes (MSS). This extends a previous analysis of perturbations done around MSS, which were shown to be ghost-free. We motivate our choice of backgrounds which arise as solutions of IDG in the UV, avoiding big bang and black hole singularities. Contrary to MSS, in this paper we show that, generically, all SVT modes are coupled to each other at the quadratic level of the action. We consider simple examples of the full IDG action, and illustrate this mixing and also a case where the action can be diagonalized and ghost-free solutions constructed. Our study is widely applicable for both non-singular cosmology and black hole physics where backgrounds depart from MSS. In appendices, we provide SVT perturbations around conformally-flat and arbitrary backgrounds which can serve as a compendium of useful results when studying SVT perturbations of various higher derivative gravity models.

gr-qc↗

arXiv2018

Non-singular metric for an electrically charged point-source in ghost-free infinite derivative gravity

In this paper we will construct a linearized metric solution for an electrically charged system in a {\it ghost-free} infinite derivative theory of gravity which is valid in the entire region of spacetime. We will show that the gravitational potential for a point-charge with mass $m$ is non-singular, the Kretschmann scalar is finite, and the metric approaches conformal-flatness in the ultraviolet regime where the non-local gravitational interaction becomes important. We will show that the metric potentials are bounded below one as long as two conditions involving the mass and the electric charge are satisfied. Furthermore, we will argue that the cosmic censorship conjecture is not required in this case. Unlike in the case of Reissner-Nordström in general relativity, where $|Q|\leq m/M_p$ has to be always satisfied, in {\it ghost-free} infinite derivative gravity $|Q|>m/M_p$ is also allowed, such as for an electron.

gr-qc↗

arXiv2018

Learning to Address Health Inequality in the United States with a Bayesian Decision Network

Life-expectancy is a complex outcome driven by genetic, socio-demographic, environmental and geographic factors. Increasing socio-economic and health disparities in the United States are propagating the longevity-gap, making it a cause for concern. Earlier studies have probed individual factors but an integrated picture to reveal quantifiable actions has been missing. There is a growing concern about a further widening of healthcare inequality caused by Artificial Intelligence (AI) due to differential access to AI-driven services. Hence, it is imperative to explore and exploit the potential of AI for illuminating biases and enabling transparent policy decisions for positive social and health impact. In this work, we reveal actionable interventions for decreasing the longevity-gap in the United States by analyzing a County-level data resource containing healthcare, socio-economic, behavioral, education and demographic features. We learn an ensemble-averaged structure, draw inferences using the joint probability distribution and extend it to a Bayesian Decision Network for identifying policy actions. We draw quantitative estimates for the impact of diversity, preventive-care quality and stable-families within the unified framework of our decision network. Finally, we make this analysis and dashboard available as an interactive web-application for enabling users and policy-makers to validate our reported findings and to explore the impact of ones beyond reported in this work.

stat.AP↗

arXiv2017

Robin Gravity

We write down a Robin boundary term for general relativity. The construction relies on the Neumann result of arXiv:1605.01603 in an essential way. This is unlike in mechanics and (polynomial) field theory, where two formulations of the Robin problem exist: one with Dirichlet as the natural limiting case, and another with Neumann.

gr-qc↗