Oxidation-State Assignment

Choosing optimized weights for the base models
Choosing optimized weights for the base models
Proxy R² ≈ 0.91 ~7k OQMD Ensemble (soft voting)

What: Assign oxidation states in MOFs with a soft-voting ensemble of four diverse base learners.

Why: Reproduced and extended a published approach (DOI) to validate best practices on a custom OQMD/ICSD subset.

How: Curated ~7k ICSD-tagged MOF entries from OQMD; manual target assignment; tuned four base models via mixed strategies (random search, simulated annealing, TPE) and applied weighted voting.

Results: Proxy R² ≈ 0.91 on ~7k samples; custom weighting (scale 1:10) and custom candidate models improved the ensemble by ~~6% vs. uniform voting + paper's ensemble.

GitHub

Choosing optimum feature-matrix size from SHAP-ranked features (all)
Choosing optimum feature-matrix size (top 47 features from 130+ features) from SHAP-ranked features (all)
Deciding on Iteration No. for tuning the hyperparameters of the base learners using Hyperopt.
Deciding on Iteration No. for tuning the hyperparameters of the base learners using Hyperopt.