arXiv subjects

Jose Mathew

Publications and source records attributed to Jose Mathew.

9 recordsLinked to original sources

arXiv2026

Intelligent Truck Matching in Full Truckload Shipments using Ping2Hex approach

Accurate truck-to-shipment matching using GPS data is foundational for full truckload supply chain visibility, enabling real-time tracking and accurate estimated time of arrival (ETA) predictions. However, missing or corrupted vehicle identifiers prevent traditional matching approaches, leaving shipments without visibility. This paper presents Intelligent Truck Matching (ITM) 2.0, a machine learning system that addresses this critical gap by formulating matching as a probabilistic ranking problem. Our approach leverages Uber H3 hexagonal spatial indexing to discretize GPS pings into route similarity features, combined with temporal information, then applies LightGBM gradient boosting with threshold-based post-processing. Through rigorous evaluation including offline model selection (SVM, XGBoost, LightGBM), comprehensive ablation studies, and production shadow testing, we demonstrate substantial gains over rule-based baselines. ITM 2.0 achieves 26 percentage point precision improvement in North America and 14 points in Europe, while doubling coverage. Deployed in production at Project44 handling full truckload shipments, the system demonstrates robustness to geocoding errors up to 1 km, multiple candidate trucks, and sparse pings.

cs.LG↗

arXiv2024

Stability analysis of power-law cosmological models

In this paper, we revisit the stability of power-law models, focusing on an alternative approach that differs significantly from the standard approaches used in studying power-law models. In the standard approach, stability is studied by reducing the system of background FRW equations to a one-dimensional system for a new background variable $X$ in terms of the number of e-foldings. However, we rewrote the equations, incorporating $H$ into the system and went on to do the calculations up to the second order. We demonstrate by computing the deviations from the power-law exact solution to second-order in time and show that power-law contraction is never an attractor in time, regardless of parameter values. Our analysis shows that while first-order corrections align with existing interpretations, second-order corrections introduce significant deviations that cannot be explained by a simple time shift that explains the first-order diverging terms. With importance, we note that in the number of e-folds, the system remains an attractor, while in cosmic time, it is unstable. We also support our claim with numerical results. This new insight has broader implications for the study of attractor behaviour of differential equation solutions and raises questions about the stability of scenarios like the ekpyrotic bounce driven by an exponential potential. Our work also hints that the different temporal variables we use might not be equivalent.

gr-qc↗

arXiv2023

Improving Slow-Roll Estimates in Starobinsky Inflation Using Analytic Hubble Parameter

Potential slow-roll parameters are widely used in inflationary cosmology to estimate the scalar and tensor perturbation amplitudes and the scalar spectral index, although the inflationary observables are fundamentally expressed in terms of the Hubble slow-roll parameters. In this work, we revisit this approximation in the context of Starobinsky inflation in the Einstein frame. Instead of approximating the Hubble slow-roll parameters through the potential, we derive them from an analytic approximate expression for the Hubble parameter obtained in the Jordan frame and mapped to the Einstein frame. We then compare the resulting analytic predictions with numerical solutions of the background equations. We show that this procedure yields a more accurate, over the relevant interval, description of the evolution of the Hubble slow-roll parameters than the conventional potential slow-roll approximation. Consequently, for the observationally relevant value $n_s = 0.9649$, the inferred number of e-foldings decreases by more than one relative to the standard estimate, with corresponding shifts in the predicted inflationary observables. Our analysis demonstrates that the usual potential slow-roll approximation can lead to systematic deviations in precision studies of inflation, and highlights the need for more reliable estimates of the Hubble slow-roll parameters in comparisons between theoretical models and observational data.

gr-qc↗

arXiv2021

Mining Points of Interest via Address Embeddings: An Unsupervised Approach

Digital maps are commonly used across the globe for exploring places that users are interested in, commonly referred to as points of interest (PoI). In online food delivery platforms, PoIs could represent any major private compounds where customers could order from such as hospitals, residential complexes, office complexes, educational institutes and hostels. In this work, we propose an end-to-end unsupervised system design for obtaining polygon representations of PoIs (PoI polygons) from address locations and address texts. We preprocess the address texts using locality names and generate embeddings for the address texts using a deep learning-based architecture, viz. RoBERTa, trained on our internal address dataset. The PoI candidates are identified by jointly clustering the anonymised customer phone GPS locations (obtained during address onboarding) and the embeddings of the address texts. The final list of PoI polygons is obtained from these PoI candidates using novel post-processing steps. This algorithm identified 74.8 % more PoIs than those obtained using the Mummidi-Krumm baseline algorithm run on our internal dataset. The proposed algorithm achieves a median area precision of 98 %, a median area recall of 8 %, and a median F-score of 0.15. In order to improve the recall of the algorithmic polygons, we post-process them using building footprint polygons from the OpenStreetMap (OSM) database. The post-processing algorithm involves reshaping the algorithmic polygon using intersecting polygons and closed private roads from the OSM database, and accounting for intersection with public roads on the OSM database. We achieve a median area recall of 70 %, a median area precision of 69 %, and a median F-score of 0.69 on these post-processed polygons.

cs.LG↗

arXiv2020

Infinitely degenerate exact Ricci-flat solutions in f(R) gravity

We obtain an infinite number of exact static, Ricci-flat spherically symmetric vacuum solutions for a class of f(R) theories of gravity. We analytically derive two exact vacuum black-hole solutions for the same class of f(R) theories. The two black-hole solutions have the event-horizon at the same point; however, their asymptotic features are different. Our results suggest that no-hair theorem may not hold for generic modified gravity theories. We discuss the implications of our work to distinguish modified gravity theories from general relativity in gravitational wave detections.

gr-qc↗

arXiv2018

Bounce inflation driven by Higgs field

In this work, we investigate a model of bounce inflation driven by the Higgs field. The Higgs field is non-minimally coupled with gravity through the Gauss-Bonnet term. We show that the Higgs field could drive a power-law contraction followed by a bounce and thereafter an inflationary phase with exit. The phases of contraction and inflation are obtained analytically. The smooth transition to the inflationary phase from contraction is obtained numerically. Further, the power-spectrum of the model is found to be consistent with the cosmological data.

astro-ph.CO↗

arXiv2017

Exact inflationary solutions in exponential gravity

We consider a modified gravity model of the form $ f(R,\phi)=R e^{h(\phi)R} $, where the strong gravity corrections are taken to all orders and $\phi$ is a self-interacting massless scalar field. We show that the conformal transformation of this model to Einstein frame leads to non-canonical kinetic term and negates the advantage of the Einstein frame. We obtain exact solutions for the background in the Jordan frame without performing conformal transformations and show that the model leads to inflation with exit. We obtain scalar and tensor power-spectrum in Jordan frame and show that the model leads to red-tilt. We discuss the implications of the same in the light of cosmological observations.

gr-qc↗

arXiv2017

Inflation with $f(R,\phi)$ in Jordan frame

We consider an $f(R)$ action that is non-minimally coupled to a massive scalar field. The model closely resembles scalar-tensor theory and by conformal transformation can be transformed to Einstein frame. To avoid the ambiguity of the frame dependence, we obtain an exact analytical solution in Jordan frame and show that the model leads to a period of accelerated expansion with an exit. Further, we compute the scalar and tensor power spectrum for the model and compare them with observations.

gr-qc↗

arXiv2016

Low Scale Higgs Inflation with Gauss-Bonnet Coupling

Recent LHC data provides precise values of coupling constants of the Higgs field, however, these measurements do not determine its coupling with gravity. We explore this freedom to see whether Higgs field non-minimally coupled to Gauss-Bonnet term in 4-dimensions can lead to inflation generating the observed density fluctuations. We obtain analytical solution for this model and that the exit of inflation (with a finite number of e-folding) demands that the energy scale of inflation is close to Electro-weak scale. We compare the scalar and tensor power spectrum of our model with PLANCK data and discuss its implications.

astro-ph.CO↗