Catalytic Nitrate Reduction Data for Pd- and Pt-Based Catalysts

A literature-curated dataset underpins a multitask Gaussian process framework that improved predictive stability by reducing the generalization gap by 41–45% in ΔR². The data, authored by Md. Mahjib Hossain and last updated in May 2026, supports the analysis of shared descriptors governing activity and selectivity in catalytic nitrate reduction. External validation and knowledge transfer to a data-limited kinetic dataset further support the framework's robustness within the studied domain.

Use Cases

• Training multitask learning models to predict catalytic activity and ammonium selectivity based on catalyst properties.
• Identifying key catalytic descriptors for nitrate reduction using feature-importance frameworks.
• Validating catalyst performance predictions against newly published experimental data.
• Transferring learned activity-selectivity relationships to data-limited kinetic datasets.

Strengths

• The multitask learning framework reduced the generalization gap by 41–45% in ΔR² and up to 48% in ΔRMSE relative to single-task models.
• The dataset is derived from a large literature-curated collection of Pd- and Pt-based catalysts.
• Predictive robustness is supported by external validation using newly published data.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset's 222.3 KB size indicates a limited scope, likely focusing on curated summary statistics rather than raw experimental records.

Provenance

Source

figshare

Collection Method

Literature-curated from studies on Pd- and Pt-based catalysts.

Freshness

Last updated 2026-05-28 20:32:04