qm9

A short and easy PyTorch implementation of E(n) Equivariant Graph Neural Networks

A Python package for data-mining the QM9 dataset

Predicting properties of small molecules using MPNN on QM9 dataset

A function-centric graph neural network approach for predicting electron densities.

Advanced Guidance Strategies for Conditional Molecule Generation with Flow Matching

Atom-JEPA: Joint-Embedding Predictive Architecture for 3D Atomistic Systems

Graph Neural Network creation module, implemented in Tensorflow 2 with examples using the module and the iGNNition library for fast GNN prototyping.

A from-scratch Message Passing Neural Network (MPNN) for molecular property prediction, built and benchmarked on QM9-style data. Classical AI/DL baseline for a quantum-enhanced drug discovery research project.

A Variational Autoencoder in Google Colab to generate and visualize novel molecular structures for potential drug discovery applications, using the QM9 dataset and SMILES representation.

Machine learning pipeline for predicting HOMO-LUMO gap of molecules using the QM9 quantum chemistry dataset (Python, scikit-learn)

Raman Spectrum Predictor: A machine learning model trained on QM9-like molecules to estimate Raman peak positions near 1600 cm⁻¹. Includes training pipeline, Streamlit UI, and RDKit-based featurization.

Hybrid quantum-classical graph neural network for molecular property prediction on QM9, built with PennyLane, PyTorch Geometric, and RDKit.

A update QM9 benchmark, update form the recent articles

Does persistent homology help an E(3)-equivariant model predict molecular dipole moments and polarizability tensors? A controlled study with matched negative controls, whose answer in this configuration is no.

Molecular property prediction on QM9 using Graph Neural Networks with PyTorch Geometric, including GCN and bond-aware NNConv models.

QM9 molecular gap prediction with SchNet, DimeNet++ and ViSNet under reconstructed geometry. Fixed splits, training diagnostics and frozen evaluation records.

A graph neural network project for predicting molecular HOMO-LUMO gap values using QM9-style molecular graph data.

Master's Thesis: Constrained Molecular Graph Generation with Diffusion Models

SO(3)-эквивариантное предсказание дипольного вектора и тензора поляризуемости молекул (QM9S). SchNet + физические головы ранга 1 и 2, верификация эквивариантности, абляция TDA-признаков с объяснением через теорию представлений.

QM9 molecular prediction with EGNN and topology features: paired evaluation, four-condition three-seed validation and a bounded negative result.