arXiv subjects

Ali Davariashtiyani

Publications and source records attributed to Ali Davariashtiyani.

4 recordsLinked to original sources

arXiv2025

HyperDiffusionFields (HyDiF): Diffusion-Guided Hypernetworks for Learning Implicit Molecular Neural Fields

We introduce HyperDiffusionFields (HyDiF), a framework that models 3D molecular conformers as continuous fields rather than discrete atomic coordinates or graphs. At the core of our approach is the Molecular Directional Field (MDF), a vector field that maps any point in space to the direction of the nearest atom of a particular type. We represent MDFs using molecule-specific neural implicit fields, which we call Molecular Neural Fields (MNFs). To enable learning across molecules and facilitate generalization, we adopt an approach where a shared hypernetwork, conditioned on a molecule, generates the weights of the given molecule's MNF. To endow the model with generative capabilities, we train the hypernetwork as a denoising diffusion model, enabling sampling in the function space of molecular fields. Our design naturally extends to a masked diffusion mechanism to support structure-conditioned generation tasks, such as molecular inpainting, by selectively noising regions of the field. Beyond generation, the localized and continuous nature of MDFs enables spatially fine-grained feature extraction for molecular property prediction, something not easily achievable with graph or point cloud based methods. Furthermore, we demonstrate that our approach scales to larger biomolecules, illustrating a promising direction for field-based molecular modeling.

cs.LG↗

arXiv2024

Impact of Data Bias on Machine Learning for Crystal Compound Synthesizability Predictions

Machine learning models are susceptible to being misled by biases in training data that emphasize incidental correlations over the intended learning task. In this study, we demonstrate the impact of data bias on the performance of a machine learning model designed to predict the synthesizability likelihood of crystal compounds. The model performs a binary classification on labeled crystal samples. Despite using the same architecture for the machine learning model, we showcase how the model's learning and prediction behavior differs once trained on distinct data. We use two data sets for illustration: a mixed-source data set that integrates experimental and computational crystal samples and a single-source data set consisting of data exclusively from one computational database. We present simple procedures to detect data bias and to evaluate its effect on the model's performance and generalization. This study reveals how inconsistent, unbalanced data can propagate bias, undermining real-world applicability even for advanced machine learning techniques.

cond-mat.mtrl-sci↗

arXiv2021

Understanding the role of anharmonic phonons in diffusion of bcc metal

Diffusion in high-temperature bcc phase of IIIB-IVB metals such as Zr, Ti, and their alloys is observed to be orders of magnitude higher than bcc metals of group VB-VIB, including Cr, Mo, and W. The underlying reason for this higher diffusion is still poorly understood. To explain this observation, we compare the first-principles-calculated parameters of monovancy-mediated diffusion between bcc Ti, Zr, and dilute Zr- Sn alloys and bcc Cr, Mo, and W. Our results indicate that strongly anharmonic vibrations promote both the vacancy concentration and the diffusive jump rate in bcc IVB metals and can explain their markedly faster diffusion compared to bcc VIB metals. Additionally, we provide an efficient approach to calculate diffusive jump rates according to the transition state theory (TST). The use of standard harmonic TST is impractical in bcc IIIB/IVB metals due to the existence of ill-defined harmonic phonons, and most studies use classical or ab initio molecular dynamics for direct simulation of diffusive jumps. Here, instead, we use a stochastically-sampled temperature-dependent phonon analysis within the transition state theory to study diffusive jumps without the need of direct molecular dynamics simulations. We validate our first-principles diffusion coefficient predictions with available experimental measurements and explain the underlying reasons for the promotion of diffusion in bcc IVB metals/alloys compared to bcc VIB metals.

cond-mat.mtrl-sci↗

arXiv2020

Phonon-assisted diffusion in bcc phase of titanium and zirconium from first-principles

Diffusion is the underlying mechanism for many complicated materials phenomena, and understanding it is basic to the discovery of novel materials with desired physical and mechanical properties. Certain groups of solid phases, such as the bcc phase of IIIB and IVB metals and their alloys, which are only stable when they reach high enough temperatures and experience anharmonic vibration entropic effects, exhibit "anomalously fast diffusion". However, the underlying reason for the observed extraordinary fast diffusion is poorly understood and due to the existence of harmonic vibration instabilities in these phases the standard models fail to predict their diffusivity. Here, we indicate that the anharmonic phonon-phonon coupling effects can accurately describe the anomalously large macroscopic diffusion coefficients in the bcc phase of IVB metals, and therefore yield a new understanding of the underlying mechanism for diffusion in these phases. We utilize temperature-dependent phonon analysis by combining ab initio molecular dynamics with lattice dynamics calculations to provide a new approach to use the transition state theory beyond the harmonic approximation. We validate the diffusivity predictions for the bcc phase of titanium and zirconium with available experimental measurements, while we show that predictions based on harmonic transition state theory severely underestimates diffusivity in these phases.

cond-mat.mtrl-sci↗