arXiv subjects

Justin Tahmassebpur

Publications and source records attributed to Justin Tahmassebpur.

4 recordsLinked to original sources

arXiv2026

A geometric basis for materials families in inorganic solids

The thermodynamic stability of inorganic solids spans a vast compositional space, yet materials scientists have long organized their intuition around a manageable number of materials families. Here we show that this organization has a precise geometric basis. The formation-energy convex hull of all inorganic compounds from the Materials Project, spanning 92-dimensional elemental composition space, is captured to near DFT accuracy by a polyhedron with only seven facets. Each facet corresponds to a family of materials sharing similar chemical potentials. This low-dimensional structure is not merely an economical description of energies: without retraining or structural input, the same framework reproduces trends in DFT-calculated defect energies and elemental spatial correlations in high-entropy nanoparticles. These results reveal that a small number of material families, corresponding to geometric features of composition-energy space, govern bulk stability, defect energetics, and elemental mixing, and provide a unified, interpretable framework for rapid screening across diverse materials systems.

cond-mat.mtrl-sci↗

arXiv2026

Learning from almost nothing: How neural networks survive heavy input corruption

Learning from imperfect data is a central theme in machine learning, connecting practical questions of robustness to fundamental questions of learnability. Here we examine attribute noise: learning from corrupted inputs while keeping the labels intact, a setting that has received considerably less analytical attention than its label-noise counterpart. We consider two types of corruption models: additive noise and replacement noise. Through experiments with multi-layer perceptrons (MLPs) on corrupted classification datasets, we find that neural networks remain robust, maintaining well-above-chance accuracy even when inputs are >90% corrupted -- far beyond human recognition. To understand this robustness, we analyze infinite-width networks in the heavy-corruption regime using a mean-field-inspired approach and derive a leading-order decision rule for the classification outcome: the network implements a prototype rule, the nearest-class-mean, assigning each test point to the class whose training-set average it most closely resembles. This leading-order decision rule is universal across a broad range of MLP architectures, holding for any depth, as well as a wide class of activation functions and noise distributions. The same centroid mechanism closely matches finite-width network behavior in our experiments and provides an interpretable and analytically tractable account of why learning can succeed even when individual training examples carry almost no signal.

cs.LG↗

arXiv2025

Effective Atom Theory: Gradient-Driven ab initio Materials Design

We introduce Effective Atom Theory (EAT), a framework that transforms combinatorial materials design into a smooth, gradient-driven optimization within density functional theory (DFT). Atoms are represented as probabilistic mixtures of elements, enabling gradient-based optimizers to converge to a physically realizable material in about 50 energy evaluations -- far fewer than combinatorial optimization methods. Applied to Co-Cr-Ni-V oxides for the alkaline oxygen evolution reaction (OER), EAT leads to a final recommended composition of Co0.19Cr0.06V0.31Ni0.44O.

cond-mat.mtrl-sci↗

arXiv2022

How Turbulent Jets Can Disperse Virus Clouds in Poorly Ventilated Spaces

We show that enhanced turbulent mixing can be used to mitigate airborne COVID-19 transmission by dispersing virus-laden clouds in enclosed, poorly ventilated spaces. A simple system of fan-driven turbulent jets is designed so as to minimize peak concentrations of passive contaminants on time scales short compared to the room ventilation time. Standard Reynolds-average and similarity methods are used, and combinations of circular and radial wall jets are considered. The turbulent diffusivity and contaminant mixing time are calculated. Results indicate that this approach can significantly reduce peak virus cloud concentrations, especially in small spaces with low occupancy, such as restrooms. Turbulent mixing is, of course, ineffective in the absence of ventilation.

physics.flu-dyn↗