Home » Tags | Hunter Heidenreich, AI Research Scientist

Generative-Models

Scientific Computing
Grid of complex molecular structures rendered from SELFIES and SMILES strings

Molecular String Renderer: Robust Visualization Infrastructure

A robust, type-safe Python library for converting chemical string representations (SMILES, SELFIES, InChI) into …...

Generative Modeling
Diagram comparing standard stochastic sampling (gradient blocked) vs the reparameterization trick (gradient flows)

Auto-Encoding Variational Bayes (VAE Paper Summary)

Summary of Kingma & Welling's foundational VAE paper introducing the reparameterization trick and variational …...

Generative Modeling
Flowchart comparing VAE and IWAE computation showing the key difference in where averaging occurs relative to the log operation

Importance Weighted Autoencoders (IWAE Paper Summary)

Summary of Burda, Grosse & Salakhutdinov's ICLR 2016 paper introducing Importance Weighted Autoencoders for tighter …...

Generative Modeling
MNIST digit samples generated from a Variational Autoencoder latent space

Importance Weighted Autoencoders: Beyond the Standard VAE

The key difference between multi-sample VAEs and IWAEs: how log-of-averages creates a tighter bound on log-likelihood.

Computational Chemistry

SELFIES: The Original Paper (Krenn et al. 2020)

The 2020 paper introducing SELFIES, the 100% robust molecular representation that solves SMILES validity problems in ML.

Computational Chemistry
SELFIES strings guarantee 100% valid molecules - even when generated randomly

Converting SELFIES Strings to 2D Molecular Images

Visualize SELFIES molecular representations and test their 100% robustness through random sampling experiments.

Computational Chemistry

DenoiseVAE: Learning Molecule-Adaptive Noise Distributions for Denoising-based 3D Molecular Pre-training

Liu et al.'s ICLR 2025 paper introducing DenoiseVAE, which learns adaptive, atom-specific noise for better molecular …...

Generative Modeling
Visualization of the VAE prior hole problem showing a ring-shaped aggregate posterior with an empty center where the Gaussian prior has highest density

A Contrastive Learning Approach for Training Variational Autoencoder Priors

Aneja et al.'s NeurIPS 2021 paper introducing Noise Contrastive Priors (NCPs) to address VAE's 'prior hole' problem with …...

Computational Chemistry

Invalid SMILES are Beneficial Rather than Detrimental to Chemical Language Models

Skinnider (2024) shows that generating invalid SMILES actually improves chemical language model performance through …...

Generative Modeling
Variational Autoencoder architecture diagram showing encoder, latent space, and decoder

Modern PyTorch Techniques for VAEs: A Hands-On Tutorial

Learn to implement VAEs in PyTorch: ELBO objective, reparameterization trick, loss scaling, and MNIST experiments on …

Generative Modeling
Wasserstein distance visualization showing Earth-Mover distance concept for GAN training

GAN Objective Functions: A Comprehensive Guide

Complete guide to GAN objective functions including WGAN, LSGAN, Fisher GAN, and more. Understand which loss function to …

Generative Modeling
Illustration of GAN training process showing adversarial competition between generator and discriminator

Understanding Generative Adversarial Networks (GANs)

Learn about GANs with intuitive explanations and mathematical foundations. Learn how adversarial networks generate …