TabPFN outperforms published models on three energetic-material properties
Original title: Accurate predictions of properties of energetic materials with tabular foundation models
Untuned TabPFN on Chemprop embeddings beats Chemprop itself and is more than ten times faster than XGBoost.
Authors apply TabPFN, a tabular foundation model, to predict impact sensitivity, decomposition temperature, and crystal density of energetic materials. Without hyperparameter tuning, they report it beats the best published models and tuned XGBoost, random forests, and Gaussian processes, whether given hand-crafted descriptors or Chemprop embeddings. TabPFN predicts a full target distribution for uncertainty estimates and is interpreted with Shapley values.
Unconfirmed: Accuracy and speed claims are the authors' comparisons, not independently verified.
- TabPFN on Chemprop embeddings is more accurate than Chemprop for all three properties.
- It is more than an order of magnitude faster than XGBoost among models tested.
- Uncertainty estimates are compared with repeated measurements of the same compounds.
- Shapley values rank which features drive the predictions.
- AI models
- TabPFN
From the original article
1 min read at arxiv.orgWe use TabPFN, a recently developed tabular foundation model, to predict three properties of energetic materials: impact sensitivity, decomposition temperature and crystal density. Without any hyperparameter tuning, TabPFN outperforms the best published models for all three properties.
By Vagn Krister Nes Glomsrud · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.