Data-Driven Design of Organic Semiconductors via Machine Learning Framework

A dataset supporting the design of organic semiconductor molecules using a hierarchical variational autoencoder, Gaussian mixture regression, and Bayesian optimization. The dataset, created by Yamato Nakanishi and last updated in June 2026, contains generated molecular candidates optimized for low reorganization energy and high carrier mobility. The framework identified a new minimum hole reorganization energy molecule via sulfur-to-nitrogen substitution.

Use Cases

• Generating novel organic semiconductor candidates based on latent space sampling.
• Predicting hole and electron reorganization energies for small molecules.
• Optimizing molecular structures for high carrier mobility using Bayesian optimization.
• Analyzing the relationship between structural descriptors and electronic transport properties.

Strengths

• Dataset size is 1.1 MB, indicating a focused and manageable collection.
• The description details a specific methodological framework combining HVAE, GMR, and Bayesian optimization.
• The work resulted in the identification of a molecule with a new minimum hole reorganization energy.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect the specific generative model bias inherent to the described framework.

Provenance

Source

figshare

Collection Method

Generated via a machine learning framework combining hierarchical variational autoencoders, Gaussian mixture regression, and Bayesian optimization.

Freshness

Last updated 2026-06-04 05:03:37.