Papers and Code

KGQA4MAT

KGQA4MAT is a novel benchmark for natural language (NL) question answering over a knowledge graph for Metal-Organic Frameworks (MOFs). The MOF knowledge graph (MOF-KG) is built by integrating structured databases and knowledge extracted from the literature. The graph is stored in a Neo4j database. To enable domain experts to ask NL questions from the MOF-KG, we develop a natural language interface leveraging ChatGPT to translate NL questions to Cypher queries. In this study, we demonstrate the potential of our approach through the benchmark dataset and the well-known QALD-9 dataset.

LLM-Prop

LLM-Prop is an efficiently finetuned large language model (T5 encoder) on crystals text descriptions to predict their properties. Given a text sequence that describes the crystal structure, LLM-Prop encodes the underlying crystal representation from its text description and output its properties such as band gap and volume.

Mixture of Experts for Materials Science

This repository contains a general framework for leveraging complementary information across different models and datasets for accurate prediction of data-scarce materials properties.

Vendi Score

The Vendi Score (VS) is a metric for evaluating diversity in machine learning. VS can be interpreted as the effective number of unique elements in a sample. It increases linearly with the number of modes in the dataset. VS is highest when the items in the sample differ in many attributes, and the attributes are not correlated with each other.

Vendi Sampling

Vendi Sampling is a method for increasing the efficiency and efficacy of the exploration of molecular conformation spaces. In Vendi sampling, molecular replicas are simulated in parallel and coupled via a global statistical measure, the Vendi Score, to enhance diversity.

YAMZ

Yet Another Metadata Zoo (YAMZ) is a collaborative vocabulary application which allows individuals to define terms as well as comment and vote on them. It allows communities to develop vocabularies specific to their needs, especially to improve the quality of metadata used. Each term receives a permanent identifier in the form of an archival key (ARK) that allows users to have a stable linked data reference.

OTMapOnto Ontology Matching Tool

This repository contains the sources of the ontology matching tool, OTMapOnto, and the results of participating in the Ontology Alignment Evaluation Initiative 2021. The tool is wrapped as a Web service.

PyRAI2MD

The Python Rapid Artificial Intelligence Ab Initio Molecular Dynamics code (PyRAI2MD) code is a machine learning software that accelerates photodynamics simulations by 100,000 times. PyRAI2MD enables long timescale simulations in complex environments such as explicit solvation and for molecules of unprecedented complexity.

Multi-fidelity-MCMC

Multi-fidelity Markov chain Monte Carlo (MCMC) algorithms combine models of varying fidelities in order to obtain an approximate target density with lower computational cost. We describe a class of asymptotically exact multi-fidelity MCMC algorithms for the setting where a sequence of models of increasing fidelity can be computed that approximates the expensive target density of interest. We take a pseudo-marginal MCMC approach for multi-fidelity inference that utilizes a cheaper, randomized-fidelity unbiased estimator of the target fidelity constructed via random truncation of a telescoping series of the low-fidelity sequence of models.

Vitruvion

Vitruvion is a generative model of parametric CAD sketches, which constitute the basic computational building blocks of modern mechanical design. Our model, trained on real-world designs from the SketchGraphs dataset, autoregressively synthesizes sketches as sequences of primitives, with initial coordinates, and constraints that reference back to the sampled primitives. As samples from the model match the constraint graph representation used in standard CAD software, they may be directly imported, solved, and edited according to downstream design tasks.

JAX FDM

JAX FDM enables the solution of inverse form-finding problems for discrete force networks using the force density method (FDM) and gradient-based optimization. It streamlines the integration of form-finding simulations into deep learning models for machine learning research.

COMPAS CEM

Inverse design of 3D trusses via automatic differentiation. COMPAS CEM encapsulates the CEM framework into an open-source structural design tool that enables the formulation and the solution of constrained form-finding problems in plain and simple Python code.

VERDE Materials Database

The Virtual Excited State Reference for the Discovery of Electronic Materials Database. We have established this resource in line with our interest in light-responsive π-conjugated organic molecules with applications in green chemistry, organic solar cells, and organic redox flow batteries. It includes results of our active and past virtual screening studies; to date, more than 13000 density functional theory (DFT) calculations have been performed on 1500 molecules to obtain frontier molecular orbitals and photophysical properties, including excitation energies, dipole moments, and redox potentials.

Material_Recommender

Material_Recommender leverages representations extracted from language models pretrained on material science literature for material discovery and property prediction. Code still under construction.