iiinova — where your digital self keeps evolving

majhas.bsky.social @majhas.bsky.social · 10/06/2025

The gifs didn't post properly 😅 Here is one showing the electron cloud in two stages: (1) the learning of electron density during training and (2) the predicted ground-state across conformations 😎

020

majhas.bsky.social @majhas.bsky.social · 10/06/2025

(9/9)⚡ Runtime efficiency Self-refining training reduces total runtime up to 4 times compared to the baseline and up to 2 times compared to the fully-supervised approach!!! Less need for large pre-generated datasets — training and sampling happen in parallel.

120

majhas.bsky.social @majhas.bsky.social · 10/06/2025

(8/n) 🧪 Robust generalization We simulate molecular dynamics using each model’s energy predictions and evaluate accuracy along the trajectory. Models trained with self-refinement stay accurate even far from the training distribution — while baselines quickly degrade.

110

majhas.bsky.social @majhas.bsky.social · 10/06/2025

(7/n) 📊 Performance under data scarcity Our method achieves low energy error with as few as 25 conformations. With 10× less data, it matches or outperforms fully supervised baselines. This is especially important in settings where labeled data is expensive or unavailable.

110

majhas.bsky.social @majhas.bsky.social · 10/06/2025

(6/n) This minimization leads to Self-Refining Training: 🔁 Use the current model to sample conformations via MCMC 📉 Use those conformations to minimize energy and update the model Everything runs asynchronously, without need for labeled data and minimal number of conformations from a dataset!

110

majhas.bsky.social @majhas.bsky.social · 10/06/2025

(5/n) To get around this, we introduce a variational upper bound on the KL between any sampling distribution q(R) and the target Boltzmann distribution. Jointly minimizing this bound wrt θ and q yields ✅ A model that predicts the ground-state solutions ✅ Samples that match the ground true density

120

majhas.bsky.social @majhas.bsky.social · 10/06/2025

(4/n) With an amortized DFT model f_θ(R), we define the density of molecular conformations as the Boltzmann distribution This isn't a typical ML setup because ❌ No samples from the density - can’t train a generative model ❌ No density - can’t sample via Monte Carlo!

110

majhas.bsky.social @majhas.bsky.social · 10/06/2025

(3/n) DFT offers a scalable solution to the Schrödinger equation but must be solved independently for each geometry by minimizing energy wrt coefficients C for a fixed basis. This presents a bottleneck for MD/sampling. We want to amortize this - train a model that generalizes across geometries R.

120

majhas.bsky.social @majhas.bsky.social · 10/06/2025

(1/n)🚨Train a model solving DFT for any geometry with almost no training data Introducing Self-Refining Training for Amortized DFT: a variational method that predicts ground-state solutions across geometries and generates its own training data! 📜 arxiv.org/abs/2506.01225 💻 github.com/majhas/self-...