arXiv subjects

Andro Sabashvili

Publications and source records attributed to Andro Sabashvili.

5 recordsLinked to original sources

arXiv2026

Simulation-Augmented Multi-Step Split Conformal Prediction for Aggregated Forecasts

We study uncertainty quantification for aggregated forecasting tasks such as annual totals and year-over-year growth rates. We propose SA-MSCP, a simulation-augmented multi-step split conformal method that generates future paths from cross-validated residuals using a block bootstrap and constructs prediction intervals from empirical quantiles. Experiments show that SA-MSCP improves empirical coverage over a simulated-path baseline for aggregated and growth-rate targets. Our results demonstrate that simulation-enhanced conformal calibration is an effective and general framework for uncertainty quantification in aggregated time-series forecasting.

cs.LG↗

arXiv2026

Conformal Prediction Algorithms for Time Series Forecasting: Methods and Benchmarking

Reliable uncertainty quantification is of critical importance in time series forecasting, yet traditional methods often rely on restrictive distributional assumptions. Conformal prediction (CP) has emerged as a promising distribution-free framework for generating prediction intervals with rigorous theoretical guarantees. However, applying CP to sequential data presents a primary challenge: the temporal dependencies inherent in time series fundamentally violate the core assumption of data exchangeability, upon which standard CP guarantees are built. This paper critically examines the main categories of algorithmic solutions designed to address this conflict. We survey and benchmark methods that relax the exchangeability assumption, those that redefine the data unit to be a collection of independent time series, approaches that explicitly model the dynamics of the prediction residuals, and online learning algorithms that adapt to distribution shifts to maintain long-run coverage. We use AutoARIMA as the base forecaster on a large-scale monthly sales dataset, evaluating marginal coverage, interval width, and the Winkler score. Our benchmark results show that multi-step split conformal prediction method meets the 90% coverage threshold and demonstrates the best efficiency.

cs.LG↗

arXiv2013

Bilayer graphene spectral function in RPA and self-consistent GW

We calculate the single-particle spectral function for doped bilayer graphene in the low energy limit, described by two parabolic bands with zero band gap and long range Coulomb interaction. Calculations are done using thermal Green's functions in both the random phase approximation (RPA) and the fully self-consistent GW approximation. RPA (in line with previous studies) yields a spectral function which apart from the Landau quasiparticle peaks shows additional coherent features interpreted as plasmarons, i.e. composite electron-plasmon excitations. In GW the plasmaron becomes incoherent and peaks are replaced by much broader features. The deviation of the quasiparticle weight and mass renormalization from their non-interacting values is small which indicates that bilayer graphene is a weakly interacting system. The electron energy loss function, $Im[-ε^{-1}_q(ω)]$ shows a sharp plasmon mode in RPA which in GW approximation becomes less coherent and thus consistent with the weaker plasmaron features in the corresponding single-particle spectral function.

cond-mat.mes-hall↗

arXiv2011

Discretized Thermal Green's Functions

We present a spectral weight conserving formalism for Fermionic thermal Green's functions that are discretized in imaginary time and thus periodic in imaginary ("Matsubara") frequency. The formalism requires a generalization of the Dyson equation and the Baym-Kadanoff-Luttinger-Ward functional for the free energy. A conformal transformation is used to analytically continue the periodized Matsubara Green's function to the continuous real axis in a way that conserves the discontinuity at t=0 of the corresponding real-time Green's function. For given discretization the method allows numerical Green's function calculations of very high precision and it appears to give a well controlled convergent approximation as we decrease the discretization interval. The ideas are tested on Dynamical Mean Field Theory calculations of the paramagnetic Hubbard model.

cond-mat.str-el↗

arXiv2010

The Dynamical Mean Field Theory phase space extension and critical properties of the finite temperature Mott transition

We consider the finite temperature metal-insulator transition in the half filled paramagnetic Hubbard model on the infinite dimensional Bethe lattice. A new method for calculating the Dynamical Mean Field Theory fixpoint surface in the phase diagram is presented and shown to be free from the convergence problems of standard forward recursion. The fixpoint equation is then analyzed using dynamical systems methods. On the fixpoint surface the eigenspectra of its Jacobian is used to characterize the hysteresis boundaries of the first order transition line and its second order critical end point. The critical point is shown to be a cusp catastrophe in the parameter space, opening a pitchfork bifurcation along the first order transition line, while the hysteresis boundaries are shown to be saddle-node bifurcations of two merging fixpoints. Using Landau theory the properties of the critical end point is determined and related to the critical eigenmode of the Jacobian. Our findings provide new insights into basic properties of this intensively studied transition.

cond-mat.str-el↗