STESH: Attention Weights from Spatial Transcriptomics Clustering Across 200 Epochs

Attention weights from 151,672 data points generated during 200 training epochs of the STESH clustering model. The dataset supports the evaluation of STESH, a method that integrates gene expression, spatial location, and histology for spatial transcriptomics analysis. It was authored by Huihui Zhang and last updated on June 1, 2026.

Use Cases

• Analyzing the evolution of attention mechanisms during model training based on the 200-epoch progression.
• Benchmarking multi-view graph convolutional network performance based on the described expression, histology, and spatial modules.
• Studying the integration of convolutional neural network features with graph-based clustering based on the described STESH methodology.

Strengths

• Contains a specific count of 151,672 attention weight data points.
• Training process is documented across 200 epochs.
• Dataset is associated with a method that reportedly outperformed ten state-of-the-art benchmarks.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The 22.2 KB file size indicates a very limited scope, likely containing summary or model parameters rather than raw experimental data.

Provenance

Source

Huihui Zhang via figshare

Collection Method

Likely generated as output from the STESH model training process.

Freshness

Last updated 2026-06-01 17:40:34; freshness should be verified.