---
license: mit
language:
- en
pretty_name: Cascadia K Matrix
size_categories:
- 100K<n<1M
configs:
- config_name: default
  data_files:
  - split: cascadia_K_matrix
    path:
    - cascadia_K_matrix.h5.part0
    - cascadia_K_matrix.h5.part1
    - cascadia_K_matrix.h5.part2
---
# Cascadia K Matrix Dataset

Single HDF5 file split for Hugging Face (hard limit: 500 GB SI per file).

This is a dataset to be used with the AutonomousOED package: [Zenodo](https://zenodo.org/records/21820020), [GitHub](https://github.com/s769/AutonomousOED)

This dataset can be used for reproducing results in the paper:

Sreeram Venkat, Stefan Henneking, and Omar Ghattas. 
"Sensor Placement for Tsunami Early Warning via Large-Scale Bayesian Optimal Experimental Design." arXiv preprint [arXiv:2604.08812](https://doi.org/10.48550/arXiv.2604.08812) (2026). To appear in Proceedings of SC'26.

Below are the instructions to reassemble the dataset before using it in the application.

## Reassemble

```bash
cat cascadia_K_matrix.h5.part* > cascadia_K_matrix.h5
```

Parts are not valid HDF5 until concatenated in order (`part0`, `part1`, ...).

## Contents

- Original file: `cascadia_K_matrix.h5`
- Dataset: `/K_matrix` — shape `(252000, 252000)`, `float64`, chunks `(420, 420)`